How Mathematical AI Exposed Hidden Vetoes in Anthropic's S-1

LONDON – November 11, 2026 – In the high-stakes theater of corporate governance, control is rarely as straightforward as a simple majority. It is hidden in the labyrinthine clauses of shareholder agreements, conditional vetoes, and tiered voting rights. For decades, deciphering these mechanics of power required armies of corporate lawyers billing by the hour. Now, it requires ten minutes and roughly twenty dollars in computing power.

Luminance, a legal technology enterprise born out of the University of Cambridge, has unveiled a new beta capability dubbed Formal Verification. By applying mathematical proof-checking to corporate contracts, the company has introduced a radical new way to stress-test legal documents for hidden risks. To prove the viability of its new tool, the firm pointed its algorithmic lens at one of the most anticipated and complex financial documents of the year: the 400-page S-1 registration statement of artificial intelligence giant Anthropic.

The results laid bare a web of governance interdependencies, mutual vetoes, and conditional voting powers that dictate who truly controls the frontier of artificial intelligence.

Mathematical Proof Meets Corporate Law

Formal verification is not a new concept, but its application to corporate law is entirely novel. Traditionally utilized in safety-critical software engineering and aerospace, formal verification uses computational logic to establish whether a system satisfies precisely defined properties. It is the mathematical equivalent of a gold standard, ensuring that a system operates exactly as intended without fatal bugs.

Luminance’s Cambridge R&D Lab has managed to bridge the gap between this rigid mathematical certainty and the notoriously ambiguous nature of legal language. Inspired by advances in AI-assisted mathematics—such as automated solutions to the Navier–Stokes Millennium Prize Problem—the platform translates relevant contract provisions into formal rules. These rules are then checked by Lean, an interactive theorem prover programming language, to ensure every logical step follows from the established model.

“We’re always looking for ways to apply the latest AI breakthroughs to the problems lawyers face,” said Dr. Graham Sills, Director of AI and Co-Founder at Luminance. “Lean-verified mathematics creates an exciting opportunity to bring formal proof into Contract Intelligence. Formal Verification gives companies a way to test how contractual rules interact and inspect the reasoning behind the result, finding gaps in their own contracts before someone else does.”

This approach marks a significant departure from the current trajectory of the legal tech industry. While competitors like Harvey AI and Ironclad have heavily integrated natural language processing and generative models to extract clauses, draft responses, and accelerate playbook-based redlining, they primarily operate on pattern recognition. Luminance is moving beyond extraction, attempting to establish computational certainty regarding how disparate clauses interact across hundreds of pages.

Dissecting the Boardroom Power Dynamics

Anthropic’s governance structure is famously intricate, explicitly designed to balance the hyper-growth demands of a company that generated $16.3 billion in revenue in the first half of 2026 with its public benefit mandate to develop AI safely. Its S-1 filing details five distinct classes of stock, separating economic ownership from voting power and board oversight.

When Luminance’s Formal Verification tool digested this corporate architecture, it mapped the relationships between the company’s board, founders, investors, employees, public shareholders, and its unique Long-Term Benefit Trust (LTBT). The mathematical model revealed that while the LTBT appears to hold dominant control, its power is structurally conditional.

The Trust holds a single Class T share, which ostensibly grants it the power to elect and remove four of Anthropic’s seven directors. However, the AI-driven proof identified a critical dependency: the Trust cannot unilaterally choose the occupant for one of those four seats, known as the "Specified Class T Director." Changing that specific appointment requires the concurrence of a committee where non-Class T directors hold a majority.

More intriguingly, the algorithmic analysis highlighted a latent power shift tied to the company's workforce. Once Class E shares—which hold no economic rights and are designated for non-cofounder employees—are issued post-IPO, those employees gain a decisive swing vote over the Specified Class T Director.

“Questions of control matter to organizations of every size, from companies and partnerships to charities, clubs, and associations. Their articles, bylaws, and agreements determine who can act, what needs approval, and where vetoes sit," noted Don Riddick, Chief Legal Officer at Luminance. "Formal Verification can help organizations understand how changes to their structure, ownership, roles, or voting rights could affect control, and identify unintended vetoes or concentrations of power.”

The analysis proved that Anthropic’s IPO changes the fundamental balance of governance rights, creating specific new dependencies that are incredibly difficult to discern by reading each clause in isolation.

The Future of Wall Street Due Diligence

The economic and operational implications of this technology are staggering. Luminance’s AI modeled and tested the governance provisions from Anthropic's massive S-1 in roughly ten minutes, utilizing just $20 in token costs. In the realm of high-stakes mergers and acquisitions, where due diligence can take weeks and cost millions, the ability to instantly stress-test a target company's capitalization table and governance structure for logical contradictions is highly disruptive.

Investors are already taking note of the shift from theoretical AI to practical enterprise value.

“As frontier models develop, investors are looking to companies that turn those advances into trusted products that drive clear enterprise value,” said Hyun Koo, Partner at March Capital. “Formal Verification tackles this problem directly... Its application to Anthropic’s S-1 shows what this means for investors: a clearer understanding of how governance provisions interact to shape control, helping them ask more informed questions and assess potential investments as a company enters the public markets.”

If an automated system can identify an unintended concentration of power or a conflicting commitment in minutes, the traditional "lawyer in the loop" model of due diligence will be forced to evolve. Instead of spending days cross-referencing definitions and voting thresholds, corporate lawyers will increasingly act as strategic validators of algorithmic outputs.

The Limits of Algorithmic Certainty

Despite the breakthrough, the transition from legal interpretation to mathematical verification is not without friction. Law, unlike software code, is inherently subjective. It relies on context, intent, commercial reasonableness, and judicial precedent.

Luminance itself acknowledged that its model relies on "simplifications," omitting certain complexities regarding the founders' personal Class A shares, founder-split rules, and specific time limits to make the formalization tractable.

Legal experts caution against equating computational logic with legal enforceability. As one seasoned corporate governance attorney noted privately, AI tools are highly effective at processing massive volumes of text and finding logical conflicts, but pure AI still struggles with context and judgment calls. A mathematical model cannot determine if a change-of-control provision violates a board's fiduciary duties under Delaware corporate law, nor can it predict how a judge might interpret the intent behind a poorly drafted mutual veto.

Furthermore, the Delaware courts are already beginning to scrutinize the organizational controls surrounding the use of artificial intelligence in corporate decision-making. If a board relies on an algorithmic model that oversimplifies a contract, the resulting strategic decisions could be vulnerable to shareholder litigation.

Nevertheless, the introduction of theorem-proving mathematics into the legal sector represents a profound strategic shift. By translating the mechanics of power into checkable code, technology is no longer just reading the rules of the corporate game—it is beginning to audit them.