IA · 18 July 2026 · 3 min read
Beyond Self-Reporting: Why Anthropic Is Pushing for Even Tougher AI Laws
In brief: In the United States, as federal legislation remains stalled, individual states are leading the charge on AI regulation. Within this fragmented environment, Anthropic is actively lobbying for stricter rules—such as third-party auditing and injunctive enforcement—drawing sharp criticism from Silicon Valley figures who view it as 'regulatory capture' designed to lock out startups.
by Team Mocchi's
Anthropic's Regulatory Pivot
In the United States, artificial intelligence regulation is not following a centralized federal path. Faced with a persistent gridlock in Congress, individual states are setting the pace. Within this fragmented landscape, Anthropic has adopted an unusual and highly debated stance for a tech giant: actively lobbying local governments to enact even tougher regulations than those currently on the books.
As reported by Wired, Cesar Fernandez, Anthropic’s head of US state and local government relations, stated that the transparency-focused safety laws passed in 2025 by states like California and New York are already outdated. According to the company, the speed of AI development is so rapid that simple "self-reporting" obligations by AI labs are no longer sufficient to guarantee public safety.
From Transparency to Independent Auditing
The shift championed by Anthropic fundamentally redefines the concept of compliance for frontier AI. While the first wave of state laws focused on forcing companies to disclose how they trained their models, the new legislative path supported by Anthropic calls for external, binding oversight.
The company is throwing its weight behind proposed bills in Illinois and Massachusetts that introduce much stricter requirements. In Illinois, for instance, a proposed measure would require AI developers to have their safety processes evaluated by independent third-party auditors. In Massachusetts, the proposed policy goes further, empowering the state’s attorney general to seek injunctive relief against non-compliant companies—effectively allowing the state to halt the deployment of a model.
This transition from self-reporting to external auditing represents a major paradigm shift. For Anthropic, the goal is to preemptively mitigate catastrophic risks—such as the potential for advanced models to assist in large-scale cyberattacks or financial crises—before these systems ever reach the public.
Silicon Valley Backlash: The Regulatory Capture Debate
While Anthropic's political maneuvers have received praise from labor unions and safety advocacy groups, they have triggered fierce backlash within Silicon Valley. Many venture capitalists and founders view this push as a calculated attempt to raise the barrier of entry for smaller competitors.
Among the loudest critics is David Sacks, a prominent investor and White House technology adviser, who accused Anthropic of running a sophisticated "regulatory capture" campaign fueled by fear-mongering. The core argument is that startups and open-source projects lack the legal and financial resources needed to comply with expensive, multi-stage third-party audits. Consequently, strict laws could freeze competition, leaving the market in the hands of a few heavily funded incumbents.
Fernandez has dismissed these criticisms, arguing that the proposed laws are carefully tailored to target only the most powerful frontier models while shielding small-scale developers. Nonetheless, as the gap between frontier capabilities and accessible open-source systems continues to narrow, the debate over how to balance safety with fair competition remains highly polarized.
Mocchi's Take
From our perspective in Italy, this political divide in the United States highlights a global shift that local businesses must not ignore. Although Europe has already established its regulatory framework with the AI Act, the US state-level fragmentation creates a compliance labyrinth for any developer exporting software or working with international partners. For Italian companies, this clash proves that the era of informal self-regulation is coming to an end worldwide. Designing custom AI systems with proactive compliance, independent auditing, and rigorous data governance from day one is no longer just an ethical choice—it is a baseline requirement for international commercial viability.