Leaked Anthropic Draft AI Regulatory Bill Is Unacceptable and Troubling

Substack · Watch/Read · (gathered 2026-10-18)

• Zvi Mowshowitz finds Anthropic’s leaked draft AI regulatory bill “unacceptable and troubling,” arguing it prioritizes the company’s business interests and political risk over genuine safety, potentially making society less protected against catastrophic AI risks than under existing law.
• The bill includes “remarkably far reaching” preemption provisions that would strip states of powers to block dangerous AI (like SB 53) and sets an “absurdly high bar” for federal intervention, requiring an “imminent risk of a catastrophic harm” (e.g., >50 deaths or >$1B damage) with explicit consideration for developer harms.
• Critics widely agree the bill is weaker than the previously proposed FRONTIER Act, notably lacking strong rulemaking, emergency shutdown authority, and government licensing for independent auditors, which could lead to a “race to the bottom” in auditing standards.
• Mowshowitz warns that Anthropic needs to “course correct, quickly,” as this bill appears designed to cement a weak federal ceiling during a “lame duck” session, failing to “meet the moment” of rapidly advancing AI capabilities and prioritizing profit over safety.