Comments
Another huge thanks to Holly Scott, Aryan Yadav, Jide Alaga, Cullen O’Keefe, Haydn Bellfield and Peter Wills. This post has been greatly improved thanks to all your helpful feedback.
This is the third post in ‘Towards a Worldwide, Watertight Windfall Clause’, a sequence I’m writing on the legal viability of the Windfall Clause in seven important common law jurisdictions. Today, I’m discussing the viability of the Clause in England & Wales.
This is a very long post, so you may not want to read everything I have written. If you want to get maximum value from this post in the shortest space of time, I have four recommendations:
| Legal question: | Jurisdiction or Governing Law? | Is this a problem for the Clause? | Is this easily fixed? |
| Are valuable remedies available for breach of the Clause? | Governing Law | Yes | No |
| Is a duty of good faith implied into the Agreement? | Governing Law | Maybe | No |
| Is the Clause a breach of national competition law? | (Usually) Jurisdiction | Maybe | Yes |
| Can the Developer easily issue new shares? | Jurisdiction | Maybe | Maybe |
| Is agreeing to the Clause a breach of directors’ duties? | Jurisdiction | No | N/A |
| Can the Counterparty give good consideration for the Agreement? | Governing Law | No | N/A |
| Can a purely donative Agreement be enforced? | Governing Law | Yes | Yes |
| Are there legal restrictions on size or nature of donations? | Jurisdiction | Maybe | Yes |
| Can dividends be paid to avoid triggering the obligation? | Jurisdiction | No | N/A |
Although English[1] and Delaware contract and company law are similar, there are several differences which could make it difficult to enforce a Windfall Clause in the United Kingdom.
I am very concerned about:
I am somewhat concerned about:
I am mildly concerned about:
I am not concerned about:
Drafters of an English Windfall Clause should take the following steps:
I have also written recommendations for improving the availability of remedies for breach of the terms of the Agreement. You can find those here.
A core assumption of this post is that the Developer will choose the law of the state in which it is domiciled as the governing law of the Agreement. This is a reasonable assumption because a Developer will likely seek to benefit from home-court advantage - the fact that its legal team and domestic courts will have a deeper understanding of their domestic law and so be better placed to predict outcomes. This is a big simplification nonetheless as it's still common for multinational corporations to enter into contracts governed by a foreign law.
Relaxing this assumption should lead to a modest positive update in favour of the viability of the Windfall Clause because, in reality, it will be possible for a Developer to shop around for the most appropriate governing law for the Agreement. This will be a particularly relevant consideration in later posts, as companies domiciled in developing countries are far likelier to choose a foreign governing law.
To help anyone who would like to explore this issue further, I have highlighted which issues relate to governing law and which relate to jurisdiction in the table above.
In short: no, they are not. Check out my post on this topic here.
English courts are characteristically hostile to good faith duties, believing them to erode contractual certainty. As such, the drafters of an English Agreement must be aware of the limitations of good faith and take special care when drafting good faith provisions to ensure that they are interpreted widely enough by the courts. My recommendations here are as follows:
English courts will rarely imply good faith into a contract and have been traditionally hostile even towards express duties, preferring to exclude such provisions or interpret them as narrowly as possible. However, this is a rapidly evolving area of the law, and the traditional approach of English courts may be changing.[2] Lower courts will now interpret an express good faith clause as requiring parties to:
2. is perhaps the most important of these in the context of the Windfall Clause. If enforceable, it would allow the Counterparty to prevent the Developer from deliberately frustrating the purpose of the Agreement by, for instance, trading away WGAI to another company in its group structure.[4]
Unfortunately, recent judicial approval of expanded good faith obligations comes almost entirely from the High Court - the Court of Appeal has considered the topic of express good faith clauses just once in the last decade. In that case, the Court of Appeal aggressively narrowed the scope of an express good faith clause, stating that it applied only to two specific areas of the contract and was not an overriding duty applicable to all the parties' obligations under the contract.[5] Given the changing state of the law since this ruling, it is difficult to say how broadly the higher courts will construe express good faith clauses in the future.
The authors of the original Windfall Clause report indicate that some of the Developer’s potential escape routes from the Clause can be foreclosed by the duty of good faith implied into American contract law.[6] I would echo this assessment. Given the colossal sums at stake in a windfall-generating scenario, the possibility of a get-out would strongly incentivise the Developer to try its luck with a carefully calculated breach of contract.[7] Meanwhile, given the unusual nature of the Agreement, the drafters are unlikely to anticipate all possible failure modes, increasing the likelihood that such a get out exists. Even setting aside the legal loopholes which might be discovered by an advanced AI system, it seems risky to assume that a team of highly-skilled lawyers couldn’t find a way for the Developer to limit its obligations. This makes the inclusion of robust good faith duties essential to enforce performance.
Unfortunately, uncertainty as to the scope of good faith obligations in English law could present a serious problem for the Clause - if the higher courts construed the Developer’s good faith duties too narrowly, it would become practically impossible to enforce the terms of the Agreement. For example, if the Supreme Court decided that a duty to act in good faith did not prevent the Developer from selling WGAI if such a sale was financially desirable for the Developer’s shareholders, the shareholders may be able to force a sale even if this would substantially frustrate the Agreement.[8] A similar outcome would result if the Supreme Court rejected a general duty of good faith altogether.
The risk of narrow interpretation motivates recommendation 1. By explaining the purpose and objectives of the Clause in the recitals, the court will be encouraged to find bad faith where the Developer has clearly acted against that purpose. Recitals are not binding in English law, however, when construing the terms in a contract, the court will have regard to ‘all the background knowledge’ which both parties would have known at the time of the contract,[9] and the inclusion of the Clause’s purpose at the head of the Agreement will make it difficult for the Developer to deny its obligations. To encourage this interpretation, the recitals might also expressly link the requirements of good faith with the incompleteness of the contract, outlining that the novelty of the Agreement means the parties will not have considered all possible contingencies and intend to rely heavily on good faith for its enforceability.
This risk also motivates recommendation 2. The reticence of English courts to interpret good faith duties expansively might be counteracted by the explicit separation of a general duty of good faith, expressed to be an organising principle in the interpretation of the contract,[10] from a set of more specific good faith clauses.[11] For example, the parties might agree that the Developer should always act with good faith when negotiating the sale or licensing of intellectual property in any of its AI systems - if the Developer went on to knowingly sell WGAI, this would then prevent the Developer from claiming that it was free to act in its commercial interest in doing so. Specific clauses such as this could be strengthened by including a description of what good faith would actual entail in this particular context. For example, the parties could stipulate that a Developer acting bona fide would always notify the Counterparty of its intention to agree to a licence or sale and inform the Counterparty of the general structure of the contract, making it hard for a Developer to claim good faith if it sold WGAI off in secret. Specific duties of good faith such as this would therefore make it hard for a Developer to argue that that it was not subject to good faith duties in possible breach scenarios. Meanwhile, the existence of a further general duty to act bona fide - distinct from these specific requirements - would encourage the court to interpret all behaviour by both parties as being subject to requirement of good faith.
Unfortunately, the above two recommendations are not magic bullets. Firstly, there remains the risk that hostile English courts will simply refuse to accept such expansive duties, as they have repeatedly done in the past. Secondly, these recommendations do little to prevent inadvertent frustration of the Agreement,[12] where the Developer unknowingly places (pre-)WGAI and any related windfall profits beyond the Counterparty’s reach.[13] Thirdly, even where the court is willing in principle to accept that the Developer could have acted mala fide, the evidential burden remains high, and it will be a challenge for the Counterparty to demonstrate bad faith, particularly if the Developer has taken steps to obfuscate its intentions.
A recognition of the outstanding issues with English good faith duties motivates recommendation 3. Express prohibitions are helpful as they act as a backstop, allowing for a remedy even where a court is unwilling or unable to find bad faith. For example, the drafters might insert a requirement that the Developer must give the Counterparty a day’s notice before it grants any IPR to a third party as consideration for the issue of shares.[14] If the Developer then attempted to transfer pre-WGAI to a third party to sidestep its obligations under the Agreement, this would allow the Counterparty to seek an injunction to prevent it from doing so. Considering possible failure modes for the Agreement and then expressly prohibiting them in this way will thus help to make the contract more robustly enforceable than it would be relying on good faith alone.
Before moving on, I should also note that issues with English good faith may justify choosing a foreign governing law for the Agreement, even if England remains the chosen jurisdiction.[15] I will not consider this further here, but I intend to discuss the topic later in the sequence.
Note - Haydn Bellfield and Shin-Shin Hua also have a paper on competition law and the Windfall Clause. Much of this section is a repeat of their findings, but they reach slightly different conclusions to mine. I encourage you to read both analyses and form your own opinions.
To avoid issues with UK competition law, the parties to an English Agreement should take the following steps:
Under UK competition law, agreements or ‘concerted practices’ between organisations which have a distortionary effect on trade within the UK are prohibited.[16] This prohibition covers more than signed contracts - competition rules are interpreted teleologically, meaning the CMA will stretch the meaning of terms in legislation if this is needed to promote competitive markets.[17] Consequently, terms like ‘agreement’ and ‘concerted practice’ can include policies promoted by representative industry bodies or unspoken arrangements which involve no more than a ‘meeting of minds’. Importantly for our purposes, this will even include a unilateral public announcement by an organisation, provided that such an announcement reduces competitors’ uncertainty about that organisation’s future commercial behavior.[18] If the CMA finds that such an arrangement constitutes a violation of competition law, the arrangement is automatically void, the guilty parties become liable to pay fines of up to 10% of total worldwide turnover,[19] and the directors of the Developer may be disqualified or receive criminal penalties.[20]
Issues with competition law might arise if the Windfall Clause is adopted by multiple AI labs and widely publicised, either by industry bodies like the Partnership on AI[21] or by the labs themselves. If this behaviour was considered a concerted practice and did not qualify for any exemptions it would be void in UK law, allowing Developers to abandon their obligations under the Agreement with impunity.[22]
Windfall Clauses as a concerted practice:
The Windfall Clause is ostensibly a bilateral agreement between the Counterparty and the Developer. However, case law indicates that widespread adoption of the Clause might be considered part of an industry-wide concerted practice between Developers if each signatory to the Agreement went on to widely publicise their signatory status and encourage other Developers to sign up. The risk in such a scenario is that the CMA would consider Developers to be deliberately reducing uncertainty about their future commercial behaviour as an implicit invitation to other Developers to engage in similar anti-competitive practices. More speculatively, a concerted practice might be found for similar reasons if an industry body like PAI heavily promoted the adoption of the Clause.[23] If the CMA took such an interpretation this would be catastrophic for the Clause because every Agreement made within or outside the UK would be void and unenforceable in English law.
Admittedly, this interpretation is a stretch. Signatories can point to immediate justifications for signing the Agreement such as improved employee relations and public goodwill, making it hard to claim that the Clause is a cynical attempt to restrict competition. Furthermore, it is not clear that industry-wide adoption would have anticompetitive effects. However, as I will discuss shortly, both the Developer and CMA might face private incentives to interpret industry-wide adoption uncharitably following the achievement of windfall profits. Furthermore, certain scenarios present a greater risk of such an interpretation. For example, industry-wide adoption following the onset of an AI race might reasonably be interpreted as anticompetitive,[24] because one Developer’s decision to sign up to a Windfall Clause could reduce competitive pressures on other Developers by reducing the funds available to the signatory for reinvestment in R&D.[25] This provides a reasonable justification for taking at least some steps to mitigate the risk that industry-wide adoption of the Clause is considered anti-competitive.
The risk of such an interpretation motivates recommendation 1. The CMA offers free, non-binding advice to companies on novel questions about the application of competition law.[26] It may be worth requesting a short-form opinion from them on the circumstances in which industry-wide adoption of the Clause would violate competition law. This would allow industry bodies and the parties to the Agreement to take steps to avoid any infringement of the regulations with a better understanding of how the CMA would interpret their behaviour.
This risk also motivates recommendation 2. Provided that industry bodies and Developers themselves steer clear of actively promoting the Clause, it seems unlikely that industry-wide adoption will be considered a concerted practice. Drafters should also take care to exclude any provisions requiring the Counterparty to promote a Developer’s signatory status, as the CMA might interpret this as a roundabout attempt by a Developer to encourage industry-wide adoption. As I highlight in the post on remedies, excluding such provisions presents a risk that the Agreement will be unenforceable for want of consideration, if the Counterparty intended to offer consideration in the form of publicity. Nonetheless, the risk that the Agreement is rendered void seems important enough that the Counterparty should explore alternative forms of consideration.
The ‘technical or economic progress’ exception:
Another possible solution which could be used to justify public promotion of the industry-wide adoption of the Clause is that, under UK competition law, an arrangement between organisations which otherwise breaches competition law is exempt provided it satisfies four specific criteria. These are:
Although industry-wide adoption of the Clause likely meets criteria 2.-4., it is not clear whether the Windfall Clause promotes technical or economic progress. The issue is that the CMA’s definition of these concepts is surprisingly narrow, only including progress which is purely economic in nature and discounting even pseudo-economic benefits like improved working conditions or better public health.[27] Recent pronouncements by the CMA suggest that it might broaden this definition in the future, but there have yet to be any enforcement decisions by the authority reflecting a change of tack.[28]
Even assuming the CMA retains its current definition, it is certainly possible to argue that the Windfall Clause promotes economic progress. For instance, its proponents might claim that the redirection of windfall profits towards scientific research organisations could spur innovation, further expanding the British economy. Yet there are also reasonable arguments to the contrary. A skeptical CMA might argue that industry-wide adoption of the Clause reduces investment capital available to AI firms, slowing the rate of technological advancement. Alternatively, they might claim that increased expenditure by top firms due to windfall distributions indirectly increases prices for AI-powered goods and services, ultimately harming the consumer. Belfield and Shua raise a further concern that signatory Developers would have an incentive to reduce output as they approach profit levels that would trigger the Clause to avoid their more onerous distribution obligations.[29] All this makes it highly unclear whether or not the CMA would uphold the Clause on economic grounds.
My uncertainty here is increased when considering the private incentives of the Developer and the CMA. Setting aside the merits of each argument, as a government body the CMA might be incentivised to take the interpretation which best responded to political exigencies following the Developer’s achievement of windfall profits. Therefore, if the UK government felt it could extract greater rents by directly taxing or nationalising a Developer with WGAI, or if it believed that an unencumbered Developer would be better positioned to promote British national interests, policymakers might pressure the CMA to deliberately misinterpret the Clause. At the same time, a successful Developer may have little incentive to fight this uncharitable interpretation. Provided the risk of criminal sanctions for its directors appears minimal, the Developers may benefit from accepting a CMA fine in return for permanent release from their Agreement obligations.[30]
In light of these risks, recommendations 3. and 4. provide a simple, low-cost way of avoiding an infringement. They do so by improving the odds that the Clause qualifies for the economic progress exemption: if the stated purpose of the Agreement is the promotion of economic progress and the Counterparty can point to projects it has already undertaken to promote a healthy British economy then it will be far more difficult for the CMA to claim that the adoption of the Clause has anticompetitive effects.
Ultimately, I find it unlikely that UK competition law will present a problem for the Windfall Clause.[31] It’s true that if industry-wide public adoption of the Windfall Clause was considered a concerted practice, and if the CMA did interpret the Clause as failing to promote economic progress, then the Agreement would be unenforceable. However, both of these arguments are somewhat speculative, and a voiding of the Agreement would require the conjunction of them both. Furthermore, there appear to be cheap and effective ways to both prevent a breach of competition law and ensure that the economic progress exemption applies.
In any event, it seems worthwhile taking steps such as these to avoid even small risks of an infringement, given the serious implications for the enforceability of the Clause of any violation. Such steps will also offer comfort to the Developer and its directors, who may be averse to enter an Agreement which presents even a small risk of fines or criminal penalties.
Note - this won’t create any problems for instituting a shares-based Clause at DeepMind. If you believe DeepMind is and always will be the sole realistic English candidate to develop WGAI, you may want to skip this section.
Before implementing a shares-based Clause, proponents of an Agreement should consider the following steps:
The current procedure for issuing new shares in a limited company is simple but requires supermajority shareholder consent. Specifically, a simple majority of shareholders must vote in favour of a fresh issue of shares and a subsequent supermajority (75%) is needed to disapply shareholders’ pre-emption rights to purchase the shares.[34] If the company wants to create and issue a new class of shares, it will also need supermajority support to create those shares.[35] Note that these are only the default rules. All of the above procedures can be varied by special resolution, by the court, or on the company’s incorporation.
Under the status quo:
Provided that DeepMind remains England’s top contender to develop WGAI, it’s highly unlikely that procedural requirements will present any barriers for the implementation of a shares-based Clause. To understand why, it’s worth briefly outlining the key obstacles to implementation:
As a wholly-owned Alphabet subsidiary, DeepMind is not held back by any of these obstacles. Their subsidiary status means that there’s minimal risk of shareholder conflict, which in turn makes organisational voting rules and current classes of share largely irrelevant, as the Developer will always be able to achieve the requisite supermajority needed to implement the Clause. This should provide comfort to proponents of a shares-based Clause.
Following an industry shakeup:
Unfortunately, the absence of obstacles to implementation outlined above is contingent on the current structure of the UK’s AI industry. There are two reasons to believe that this structure may change:
These possibilities motivate recommendations 1.-3. Engagement with founders and a greater awareness of DeepMind’s (and other Developers’) constitutional structure would help ensure that it is possible to implement a shares-based Clause even if industry circumstances change. At the same time, engagement with investors would increase the likelihood of achieving sufficient support for the Clause to pass the resolutions needed to implement it. Whilst DeepMind retains its status as a wholly-controlled Alphabet subsidiary and England’s AI hegemon, there is little value in taking these steps. However, these steps may prove extremely important if the country’s AI industry experiences any significant structural shifts.
A note on the cash-based Clause:
One final thing to note is that, if a shares-based Clause was equally as viable as a cash-based Clause the above recommendations might be unnecessary, as a Developer unable to implement the former could simply opt for the latter. However, as I discussed earlier in the sequence, a shares-based Clause may offer significant advantages over a cash-based Clause in terms of available remedies. This provides additional justification for the recommendations I outlined above, as the parties may not have the option of falling back on cash distributions.[39]
It’s highly unlikely that signing the Agreement will be a breach of the directors’ duty to act in the best interests of the Developer, because English courts are generally reluctant to interfere with businesses’ freely-made commercial decisions. The original report dedicates considerable space to discussing the topic of business judgment,[40] so it seems worth explaining why the same issues are unlikely to apply in English law. This is because English courts take a passive approach to questions of business governance,[41] simply requiring directors to act in the way they honestly believe will promote the success of the company.[42] The decisions of a board cannot be appealed ‘on merits’,[43] and a court will only overrule a decision by a director if it is clearly taken in bad faith or it falls below the ‘Wednesdbury unreasonableness’ standard, meaning it was so unreasonable that no reasonable director could have considered it to be in the company’s best interests.[44]
For the directors of a Developer to be in breach of their duty in signing the Agreement, they would need to fall below this standard. This appears highly unlikely because there are good commercial reasons why a Developer might agree to the Clause, already outlined in FHI’s report.[45] To reproduce their argument in brief: a board can justify agreeing to the Clause on the grounds that it generates goodwill, attracts talented researchers with ethical concerns about advanced AI, and reduces political risk, in exchange for a very small chance of a payout. Despite the large ex post costs of the Clause for a successful Developer, this is a reasonable tradeoff for a board to make ex ante. Directors of a successful Developer who signed the Agreement can thus claim that they were acting in the best interests of the company at the time and, as such, did not breach their duties.[46]
This argument is particularly compelling in the English context, given the low bar for business judgment set by Wednesbury. Even if signing the Agreement is not an ideal business decision, it seems unlikely that a court would determine that no reasonable director could have agreed to it. It’s also worth highlighting the advantages of the English definition of ‘success of the company’, which doesn’t just consider shareholder value but requires directors to have regard to issues like the long-term consequences of their decisions and their company’s environmental impact.[47] This broad definition indicates that directors of a Developer could justify signing an Agreement for pro-social reasons, even if they could not do so on a purely economic justification. It tandem with the low requirements set by Wednesdbury, this strongly indicates that the directors of a Developer will not breach their duties to the company in signing the Agreement.
English law is famously hands-off when it comes to questions of consideration, and courts will readily uphold contracts so long as something of value has been provided by both sides.[48] This is true no matter how one-sided the contract turns out to be.[49] Given this, there is no reason to be concerned that the Windfall Clause will fail for want of consideration. Assuming the Counterparty provides something of value in return, such as a small payment or grantmaking support for other charitable donations made by the Developer, then it will be able to enforce the terms of the Agreement.[50]
Short answer: yes. Slightly longer answer: although promissory estoppel is not an independent cause of action in English law, a valid deed is enforceable in the absence of consideration.[51] An Agreement could thus be executed as a deed if the drafters wanted to avoid issues with consideration.[52]
Note - this is the least-researched part of this post, because it didn’t seem that important. Still, it’s possible I’m misinterpreting the rules around corporate donations or charity law so I’d appreciate feedback from anyone with more experience of the topic.[53]
Corporate donations are largely unregulated in the UK. Most importantly for our purposes, there do not appear to be any limitations on the size of donations which a company may make, although it’s plausible that absurdly large donations could be a breach of a board’s duty to promote the success of the company. As such, I do not anticipate that English charity regulations will be an issue for the viability of the Clause.
That said, there are two issues which are worth flagging here. I have outlined each one in turn, and each is accompanied by a simple solution:
Provided the Counterparty takes these precautionary steps, I don’t foresee any serious issues for the Agreement arising from charity regulations.
Distributions to shareholders of English companies may only be made out of profits.[58] This means that a Developer could not pay dividends or make other distributions to shareholders before meeting its obligations under the Clause. For the avoidance of doubt, the drafters might define ‘Profits’ in the contract as the Developer’s net profits before any distributions to shareholders.[59] However, it is highly unlikely that the Developer will be able to sidestep its obligations even in the absence of such a provision.
Overall, I feel less confident than I was before this project about the viability of an English Windfall Clause. On the one hand, most of the topics I’ve investigated present no issues for the viability of the Clause, and I suspect that clever drafting and careful moderation of the Counterparty’s behaviour would be enough to sidestep those issues that remain. On the other hand, I remain concerned about the hostility of English courts to good faith duties and I fear it may be impossible to achieve certainty on this issue until the Counterparty is forced to bring a case. I’m also deeply troubled by the lack of practical remedies for breach of the terms of the Agreement, which appear totally inadequate to prevent a determined Developer from breaking its promises.
It's important to stress that these remaining concerns relate primarily to English contract law, not English company law or financial regulations. This has two key implications:
In the future, I would like to see other EA lawyers explore some of these topics in more depth. In particular, I would welcome further research into how governing laws and jurisdictions might be mixed-and-matched to create a more enforceable Agreement, how the Agreement’s drafters might provide adequate contractual remedies, and a further exploration of the extent of good faith obligations in English law. In any case, I hope that this post serves as a useful jumping off point for discussion on this important topic amongst EA interested in law.
At the start of this post, I told you that you could skip most of it. That begs the question - why didn't I just cut the fat? To explain, this post is so comprehensive because I suspect it will be the most important post in the sequence. Here's why:
In this post, where I refer to ‘English law’, I am referring to the law of England & Wales. For those unfamiliar with the British legal ecosystem: English and Welsh law is practically identical and is considered a single jurisdiction. Northern Irish law is somewhat different from English law and Scottish law is a different beast entirely.
For judicial consideration of this topic, see Yam Seng Pte Ltd v International Trade Corporation Ltd (2013) EWHC 111, in particular Legatt J at [123]-[130]. This case also highlights that courts will sometimes imply good faith duties into certain contracts, such as long-term ‘relational’ contracts or contracts where there is a clear fiduciary duty between the parties. The class of contracts to which such duties apply is expanding incrementally and may eventually encompass agreements like that containing the Windfall Clause, although it does not currently do so. See also Interfoto Picture Library Ltd v Stiletto Visual Programmes Ltd [1989] 1 QB 433, in particular Bingham LJ’s comment that 'English law has, characteristically, committed itself to no such overriding principle [of good faith] but has developed piecemeal solutions in response to demonstrated problems of unfairness.'
Unwin v Bond [2020] EWHC 1768 (Comm) [229]-[230]. See also [215]-[229] for a summary of recent case law on the question of good faith. Note that this case was in the context of a shareholders’ agreement - arguably a unique class of contract in English law. It is possible that higher courts will distinguish Unwin on these grounds when dealing with future cases involving duties of good faith.
I’m encouraged by Berkeley Community Villages Ltd v Pullen [2007] EWHC 1330 (Ch), in which the court invoked a good faith clause to prevent one party from an early sale of a property which would have prevented the other party from receiving commission on the sale. One class of failure mode of a Windfall Clause involves the Developer trading away WGAI soon before it does, in fact, generate windfall profits. The Counterparty might employ similar reasoning to Berkeley to prevent this. On the other hand, there’s a good chance that the court would distinguish a Windfall case from Berkeley because the former involves a single payment, whereas the Agreement concerns a series of ongoing payments.
See Compass Group UK and Ireland Ltd (t/a Medirest) v Mid Essex Hospital Services NHS Trust [2013] EWCA Civ 200 [105]-[107].
See O'Keefe, C., Cihon, P., Garfinkel, B., Flynn, C., Leung, J. and Dafoe, A., 2020. The Windfall Clause - Distributing the Benefits of AI for the Common Good, p. 22.
I’ve considered this in more detail here.
This seems plausible. For example, a Developer, its shareholders and a third party might enter into a tripartite contract whereby the Developer sells WGAI to the third party in return for regular payments which fall just below the level that would trigger the Clause, with excess funds going directly to the shareholders. Both the shareholders and the third party could make money here, at the expense of the Counterparty, because the distributions which would have been lost to the Counterparty would be shared between them.
Investors Compensation Scheme Ltd v West Bromwich Building Society [1998] 1 WLR 896, HL(E), [912]-[913].
The idea of a ‘general organisational principle’ reflects the Canadian position since Bhasin v. Hrynew, 2014 SCC 71. In fact, it might be worth stating in the something in the contract like: ‘the general good faith duty outlined in clause X is to be interpreted in line with Canadian jurisprudence [or the jurisprudence of whichever common law country takes the most expansive interpretation of good faith]’ to make it extremely clear how this duty is to be understood.
See supra note 5. Jackson LJ commented obiter in this case that he would have been willing to construe ‘a general duty to co-operate with one another in good faith’ more broadly. This indicates that the higher courts will uphold general duties of good faith provided they are clearly indicated to be general.
Note that I am not actually certain that the contract would be ‘frustrated’ by assigning or licensing WGAI, and I've not explored this in detail. I expect the contract would only be truly frustrated if there was some reason that the Developer could never develop another WGAI after the first one. If you don't think the contract would be frustrated, replace the term with something like 'frustrated the contract, or ensured that the Windfall Clause would never be triggered'.
As I outlined in my previous post, well-informed shareholders (or other actors inside the Developer) may push an unwitting Developer to sidestep its obligations under the Clause, and it will be difficult to show that these actions were taken in bad faith.
Note that a Developer might be unwilling to agree to this if the Counterparty didn’t also agree to strict confidentiality requirements, as such a clause would also allow observers within the Counterparty to determine whether the Developer was involved in a merger.
For example, the Canadian Supreme Court recently affirmed the principle that good faith is a general organising principle in the interpretation of contractual duties (C.M. Callow Inc. v. Zollinger 2020 SCC 45). Provided that English courts applied Canadian laws consistently, an English Agreement governed by Canadian law might prove far more reliable than an attempt to shoehorn good faith into English law.
See s2 Competition Act 1998 (UK): ‘agreements between undertakings, decisions by associations of undertakings or concerted practices which may affect trade within the United Kingdom, and have as their object or effect the prevention, restriction or distortion of competition within the United Kingdom, are prohibited’.
For a helpful summary of British courts’ interpretation of EU-derived law like the Competition Act, see Interpretation of EU Legislation | Legal Guidance | LexisNexis. [online] Available at: https://www.lexisnexis.co.uk/legal/guidance/interpretation-of-eu-legislation [Accessed 1 May 2022].
See Case COMP/39850, Container Shipping, Commission Decision of 31 August 2016. In this case, the Commission determined that announcing future price increases breached Art 101(1) (the European counterpart of s2) even in the absence of direct agreement or covert contact between the parties. The Commission argued that public announcements can signal the intended conduct of each parties which in turn decreases their incentives to compete. It's possible that industry-leading Developers publicly announcing their signatory status could be construed in the same way, as substantial windfall donations could restrict the funds available for reinvestment in research, making it easier for other firms to compete. I explore this in more detail later in the post. That said, see C.f. Case T- 41/96 Bayer AG v Commission and Case C- 2/01 P BAI and Commission v Bayer: truly unilateral behaviour in the absence of any ‘concurrence of wills’ will not contravene Chapter I.
To clarify, a breach of s2(1) won’t render an entire agreement void provided that the courts can sever the offending part of the agreement. Unfortunately, this won’t help in the case of the Clause, because the offending part is not a specific clause but the very decision to sign up to the Agreement.
Parts 6-7 Enterprise Act 2002 (UK)
I think PAI would be caught by s2(1) even though it isn’t a trade association per se, because it represents the interests of many of the biggest players in AI but I haven’t looked into it. Please correct me if I am wrong here.
An ‘agreement to agree’ - assurance by the Counterparty that others Developers will also volunteer to be bound by the Windfall Clause to ensure a level playing field - may also be caught by s2: see Hua, S. and Belfield, H., 2020. AI & Antitrust: Reconciling Tensions between Competition Law and Cooperative AI Development, Yale Journal of Law and Technology 23, pp. 484-489.I won’t discuss this here, as that paper already considers the issue at length.
This point is far less certain, as there are no analogous decisions to the Container Shipping case where the European Commission or CMA has found a concerted practice resulting from a policy promoted by an industry body. Nonetheless, the CMA’s teleological approach to competition law indicates that it could still make such a finding if it believed that the Clause was anticompetitive.
This claim might be incorrect. All else being equal, an increase in the Developer’s liabilities caused by the need to make distributions under the Clause should reduce the funds it has available for other purposes, including R&D, which would correspondingly reduce competitive pressure on other Developers racing to develop WGAI. However, my analysis might be too simplistic here. For one, this would not cause any direct reduction in R&D funding as distributions would be made out of profits after accounting for reinvestment in research. Furthermore, it seems plausible that a Developer paying out some percentage of its profits under the Clause could reduce its dividend payments whilst maintaining R&D spending, which would not substantially impact competition. Overall, I’m not familiar enough with the CMA’s approach to reach a comfortable conclusion as to how it would interpret the Clause in a race scenario.
This was true at the time of writing (March 2022), but I can no longer find information about this on the CMA’s website. The authority is currently undergoing structural changes following Brexit and this service may be temporarily unavailable, or have been shelved entirely. It’s also possible that responsibility for this service is being passed to the Digital Markets Unit.
Note that this is a different approach to the European Commission. For example, see GlaxoSmithKline (2009); Rural broadband wayleave rates (2012); Modeling Sector (2016). For an in-depth analysis of the divergence in approach between the domestic competition authorities of the EU, see also Brook, O., 2019. Struggling with Article 101(3) TFEU: Diverging Approaches of the Commission, EU Courts, and Five Competition Authorities, Common Market Law Review 56 at pp. 121-156.
See Competition and Markets Authority. Retained Horizontal Block Exemption Regulations - consultation document, 2022. If the CMA starts to widen the interpretation of ‘progress’ to include sustainability benefits, this might pave the way for a further broadening of the definition to include other forms of social benefit. This seems increasingly likely given the drive to maintain compatibility of UK competition law with its EU counterpart, which will be introducing an explicit exemption to Art 101(1) TFEU for agreements which promote sustainability.
In the interest of avoiding the conjunction fallacy here, it's worth highlighting that this is a highly speculative scenario. It would only present a risk if some combination of the following occurred:
See supra note 22 at pp. 439-443, 483-489. Hua and Belfield rightly identify that the extent of this disincentive effect will depend heavily on the exact structure of the Clause. However, this effect is unlikely to disappear no matter the structure of the Clause, as the nature of the contract is to reduce the Developer’s profits.
Note that CMA fines are capped at 10% of an undertaking's worldwide turnover. Although this could mean the Developer would receive a fine greater than its obligations under the Clause in any one year, it would benefit in the short-medium term from being released from the Agreement.
To be clear, this does not mean that a Developer in possession of WGAI would not face risks from UK competition law. For example, if such a Developer held a monopoly position, the CMA might try to promote competition by breaking the business up or forcing it to license out its software to other Developers below market rate. As these sort of risks do not strictly relate to the Clause they are beyond the scope of this sequence, and so I will not consider them further here.
This is public information. For example, you can order a copy of DeepMind’s Articles of Association here.
Of course, founders might not want to implement a weird-looking Article that could discourage investment by VCs. I’m not familiar enough with startup culture to have good intuitions around whether this will be a problem, though the success of Founders' Pledge indicates that it isn't necessarily so.
ss551 and 571 Companies Act 2006 (UK). Note that the rules differ in a private limited company with only one class of shares.
s21(1) Companies Act 2006 (UK), Art 22 of the Model Articles for Private Companies Limited by Shares, and Art 43 of the Model Articles for Public Companies.
For instance, it’s plausible that some Developers will already have provisions allowing them to issue non-participating preference shares or share options. Depending on the structure of the Clause, this could do the trick.
Note that a supermajority is still needed if the Developer’s constitution requires a special resolution to disapply shareholders’ rights of pre-emption, to avoid the risk that pre-existing shareholders purchase the windfall shares.
I’m aware that the startup ecosystem is incomparably larger in the US and founder outreach efforts are likely best focused in Silicon Valley. Nonetheless, the UK is the fourth-largest producer of unicorns, home to four of the top 10 best universities, has a world-leading Office for AI focused on improving infrastructure for AI firms, and is home to DeepMind. Given this, it’s at least plausible that another DeepMind-quality Developer will emerge in the UK over the next 20-30, though I have <10% credence in this assertion.
If drafters are unsure about the enforceability of either a cash- or shares-based Clause, then this might give reason to draft a two-tiered ‘if not shares, then cash’ Clause. The idea here is that, if a shares-based Clause was found to be unenforceable or difficult to implement for any reason, the parties could fall back on a system of cash-distributions, which may improve the overall enforceability of the contract. On the other hand, a two-tiered Clause may reduce enforceability if it encourages courts to provide damages in lieu of specific performance, as damages for breach of the Clause are not likely to be sufficient.
See supra note 6 at pp.6-7.
See Carlen v Drury (1812) 1 Ves & B 154, 35 ER 61. In particular, see Lord Eldon at 63: 'The Court is not to be required on every Occasion to take the Management of every Playhouse and Brewhouse in the Kingdom.'
s172 of the Companies Act 2006 (UK) sets out this particular duty. However, this is only a codification of the common law position as per s170 of the same Act.
See Howard Smith Ltd v Ampol Petroleum Ltd (1974) 1 All ER 1126. Specifically, see Lord Wilberforce at 1131: 'There is no appeal on merits from management decisions to courts of law: nor will courts of law assume to act as a kind of supervisory board over decisions within the powers of management honestly arrived at'.
For an elaboration on the requirements placed on directors, see P Davies and S Worthington, Gower and Davies Principles of Modern Company Law (Sweet & Maxwell, 9th ed, 2012) at [16-76]. Some academics have argued for merits-based review of business decisions: for example, see Lim, Ernest, Judicial Intervention in Directors’ Decision-Making Process: Section 172 of the Companies Act 2006 (December 1, 2017). Journal of Business Law 169, 2018. This is an ongoing debate in the literature, but the current state of the case law indicates that courts are unwilling to conduct substantive review.
See supra note 6 at pp.6-7.
In this section I have mostly focused on the explanations given at pp. 6-7 of the original report, but its authors also argue at pp. 15-16 that windfall distributions could be justified in a similar manner to share options granted to startup founders. They claim that such options are justified by their low expected value at the time they are issued, even though they may have become extremely valuable by the time they are exercised. I have excluded this argument because I think there’s a significant difference between share options and promised windfall distributions. Specifically, options present an additional benefit to the company in the form of a solution to the principal-agent problem faced by investors, aligning the financial incentives of the company and its directors. As options can be justified ex ante even if they disproportionately compensate the board. Windfall distributions don’t have this same incentive effect, meaning that the directors of a Developer may need stronger arguments to justify entering into an Agreement which involves a similar financial commitment. Thanks to Peter Wills for highlighting this in his comments on drafts of this post.
See s172(1) Companies Act 2006 (UK) for a summary of the factors to which a director must ‘have regard’. As was the position at common law, these are not imperatives but rather issues which a director must consider when taking decisions. In practice, this has had little impact on how English companies are operated, but they would provide helpful cover for directors of a signatory Developer accused of breaching their duties to the company.
See Chappell and Co v Nestle Ltd (1959) UKHL 1, in which three discarded chocolate wrappers were held to be good consideration. Lord Somervell, at 114: ‘A contracting party can stipulate what consideration he chooses. A peppercorn does not cease to be good consideration if it is established that the promisee does not like pepper and will throw away the corn.’ Not everything will be good consideration: see White v Bluett (1853) 23 LJ Ex 36
See Thomas v Thomas (1842), 2 QB 851, in which £1/yr was held to be good consideration for a life interest in a house. Unfortunately, one-sided contracts can have other negative effects on the enforceability of an agreement.
Remember that, for reasons already outlined in the section on competition law, it may not be advisable for the Counterparty to provide publicity as consideration.
See Central London Property Trust Ltd v High Trees House Ltd [1947] KB 130 pp. 134-136. This case established that promissory estoppel in English law is ‘a shield, not a sword’.
An added benefit of a purely donative agreement executed as a deed is that this would avoid the risk that the Clause was considered a ‘tainted charitable donation’. Unlike ordinary donations, tainted donations are not eligible for tax relief which may make the Clause less desirable for a Developer. On the other hand, specific performance is unavailable if consideration has not moved from the injured party, which militates against the use of a deed alone. With that in mind, the best solution is likely to be that the Counterparty provides consideration and the Agreement is executed as a deed.
For anyone interested in investigationg for themselves: the law of charitable purposes is messy and inconsistent, a mix of 300 year-old case law, modern statute, and guidance by the Charity Commission. This makes the exact legal position of politically-minded charities unclear on a casual investigation. Beware of relying on old cases, even if your lexis of choice says they are still good law.
Ch. 14, Companies Act 2006 (UK)
It’s possible that the risk of political contributions is stronger than I am anticipating here. Arguably, given the sheer size of the Counterparty’s endowment, it could be difficult for the organisation to avoid straying into politics.
National Anti-Vivisection Society v Inland Revenue Commissioners [1947] UKHL 4; McGovern v Attorney-General [1982] Ch. 321.
Charity Commission guidance: Campaigning and political activity guidance for charities (CC9) at pp12-13
s830 Companies Act 2006 (UK)
Note that a Developer might use other creative accounting strategies to reduce profits or make de facto distributions to shareholders. For instance, a Developer might assign the intellectual property rights in its WGAI systems to a separate entity owned by the same shareholders but which is not bound by the Clause. It could then license back the IPR at an extremely high price, reducing the Developer’s net profits to zero and distributing any windfall gains to the same individuals. Considering such outcomes in more detail is beyond the scope of the sequence. Nonetheless, scenarios like this make it particularly important to include well-drafted good faith provisions along with other protective mechanisms I have mentioned in this post.
See Cuniberti, G., The International Market for Contracts: The Most Attractive Contract Laws, 34 Nw. J. Int'l L. & Bus. 455 (2014).
The Privy Council is a non-UK court staffed entirely by UK Supreme Court judges. It was essentially a Supreme Court for the Commonwealth, and it continues to be for many Commonwealth nations. It does not adjudicate on British cases, but its decisions are de facto binding here the UK as they are reflective of the current thinking of Britain’s most senior judges.
For an empirical analysis of the outsized persuasive value of English judgments, see Hoadley, D., Bartolo, M., Chesterman, R., Faus, A., Hernandez, W., Kultys, B., Moore, A., Nemsic, E., Roche, N., Shangguan, J., Steer, B., Tylinski, K. and West, N., 2021. A Global Community of Courts? Modelling the Use of Persuasive Authority as a Complex Network. Frontiers in Physics, 9.
For some reason the Forum isn't letting me update the post directly, so I want to highlight another core assumption which I didn't make explicit in the original post. Once the Forum starts working again I'll slot this into the post itself.
Core Assumption 2 - the Rule of Law holds together:
Later in the sequence, I'm planning to consider how a deterioration in the Rule of Law following the development of WGAI might impact the viability of the Clause. This could vary considerably by jurisdiction. For example, English constitutional law allows the legislature to break its own rules[1] if it wants to, giving Britain a unique ability amongst potential Developer host states to render the Clause inert simply by legislating that the Agreement was unlawful, or that the Developer's assets are property of the Crown.
For the moment, however, I am assuming that the Rule of Law will be largely untouched by the development of WGAI. I am doing this because it is important to explore how things might play out in a best-case scenario, where all of the relevant actors decide to play by the book. The conclusions in my post can then inform a broader analysis of the viability of the Clause in scenarios where actors' behaviour is further from the ideal.
CTRL+F 'parliament had the power to make any law except any law that bound its successors' to see Wikipedia's summary of this topic.