Almost a quarter of the paragraphs in the more than 280-page public offering memorandum of Anthropic are devoted in one way or another to warning potential future shareholders about the regulatory risk the company must face in the coming years. A careful read, for the brave, of the memorandum allows us to understand some essential aspects about the nature of regulatory risk.
First, despite the repetition of phrases about possible regulatory risks, the memorandum shows that in the 30th anniversary of the implementation of the Caremark doctrine on responsibility , the management of regulatory risk falls within the day-to-day business-as-usual activities of companies, delegated to intermediate committees or management positions that often have no voice and no vote on the board and in the design of the company’s competitive strategy, in what is a predominant culture of silence in most company executive teams when it comes to sharing with shareholders their regulatory risk management models. Despite warnings in this regard from the SEC or from the ESMA, few companies treat regulatory risk as a correctly quantified material financial risk, and not as a generic paragraph in the annual report, and Anthropic in its memorandum shows that it is not an exception to this lack of valuation of regulatory risk, even when some analysts point out that upcoming restrictions in AI regulations could imply a reduction of up to 15 percent of the company’s expected annual operating margin.
Few companies treat regulatory risk as a correctly quantified material financial risk
Anthropic shows in its memorandum (and in the actions of its top leadership) that it is yet another company in the long list that engages in reckless regulatory arbitrage when trying to design a business model based on the assumptions of an incomplete or sometimes nonexistent regulatory framework, characteristic, for example, of new market creators such as Airbnb, SpaceX, Uber or now OpenAI or Anthropic, without there being a Plan B when the different regulators close the gap. Although many executives are still not aware of this, it has been many years since the “Harvard” and static view of the competitive advantage based either on the offering of differentiated products or on lower costs has been surpassed.
The new vision of business leadership, more “siliconvalley-like,” understands that the objective of current management consists in creating a competitive advantage from dynamic capabilities, which are unique and differential organizational routines in the areas in which the company can detect and exploit business opportunities. Within these routines that lead to competitive advantage, companies must obligatorily build routines for detecting and quantifying regulatory risk, optimizing resources and processes based on the regulatory risk that has been detected, and constructing mechanisms that provide the company with sufficient flexibility to adapt quickly to regulatory risks. Despite the thousands of words devoted to regulatory risk, there is no trace of the description of these routines in Anthropic. In the artificial intelligence sector the question is not whether it will generate new regulation; the question is simply when the same will be approved. Faced with an environment of profound regulatory change, Anthropic does not explain what routines/processes it has put in place to be able to leverage the regulatory change that is yet to come in order to maintain or improve its competitive position. Rather, on the contrary, in the Anthropic brochure it acknowledges that some actions or statements by the company and its executives may have put it at a disadvantage compared to other competitors in terms of the impact of regulatory change.
Another of the major shortcomings identified in the brochure is that clearly the company (or the lawyers who drafted the document) does not understand the process by which regulatory risk is generated. Regulation (in democratic countries, at least) is usually created to respond to demands to defend the general interest. The materialization of the general interest in specific demands from citizens and their politicians does not usually happen overnight, but rather it often needs several years of slow cooking and alignment of dispersed citizen demands and political will that is not very clearly defined.
Companies must design proactive strategies in the face of this regulatory change process, trying to incorporate into it as soon as possible and not waiting for the later phases of the process such as the phase of parliamentary processing or judicialization. Intervening too late in the consolidation process of the general interest reduces companies’ strategic options and can imply very relevant costs, as Anthropic and the $1.5 billion fine paid in the case Bartz vs Anthropic teach us again, which reflects how the company, like other companies in the sector, was negligent when it came to building a sufficient base of social legitimacy that would allow the construction of the concept of acceptable use based on the provenance of the training data of the now dominant AI models in society to be close to its interests.
In the Anthropic brochure, it recognizes that some actions or statements by the company and its executives may have placed it at a disadvantage compared with other competitors in terms of the impact of the regulatory change
Another characteristic of the process of building regulatory risk is the existence of very strong bilateral information asymmetries. On the one hand, regulators (in some countries more than in others) often lack sufficient information to be able to legislate correctly about a market. As with other American companies in the sector that try to set up barriers to entry to Chinese technology, in the Anthropic brochure it states that regulation should define the speed of technological change, in such a way that AI capabilities grow in a security environment. This statement, which really reveals the weakness of Antropic’s business model in the face of low-end disruptive strategies, does not take into account that in order to align regulation with AI development, information asymmetries would have to be eliminated in such a way that Anthropic and the other companies in the sector would have to share very sensitive information with the regulator and with society, something that would blow up its technological advantage.
The other side of the asymmetries of information consists in the fact that since Anthropic does not track the process of change in regulatory risk from its earliest phases, it is unable to foresee the most likely trends in it: this complete ignorance is reflected in the brochure, full of generalities about regulatory change, but with hardly any concreteness. For example, an appropriate regulatory change impact analysis for a company must be carried out for the entire value chain. Well, since the company is betting in the coming years on the creation of multiple vertical AI markets, Anthropic makes some reference to regulatory risk in the vertical markets of its clients (finance, health, etc.), but ignores the regulatory risk derived from suppliers of resources and capabilities, such as electricity generation and distribution markets, water supply, or component or raw materials markets subject to regulatory change linked to the concept of technological sovereignty prevailing within the framework of the general interest of many countries. In this incomplete description of the value chain, there is also no mention of the sector’s competitive dynamics, based on platforms, which usually involves much closer monitoring and stricter regulation by the guarantors of competition in the markets.
Anthropic’s initial public offering exposes the persistent disconnect between the rhetoric about regulatory risk and its true integration into the company’s competitive strategy. Despite devoting hundreds of pages to formal warnings, the company addresses regulation from a defensive and reactive posture: it relegates its management to mere compliance, lacks dynamic capabilities to quantify its real impact on the value chain and reinforce a competitive advantage over the construction of the general interest. In a disruptive sector where regulatory intervention is a certainty, the real competitive advantage in the digital era does not depend solely on technological sophistication, but on the company’s ability to integrate regulatory intelligence and social legitimacy directly into the core of the strategy. Treating it as an exogenous element to the strategy is tantamount to negligently assuming a top-tier strategic vulnerability.