Norms Impact
Anthropic ditches its core safety promise in the middle of an AI red line fight with the Pentagon | CNN Business
A federal contracting ultimatum collides with a private company’s retreat from hard AI safety commitments, normalizing government leverage over guardrails that protect citizens from surveillance and misuse.
Feb 25, 2026
⚖ Legal Exposure
Sources
Summary
Anthropic replaced its prior Responsible Scaling Policy with a nonbinding safety framework and removed its commitment to pause training when model capabilities outstrip the company’s ability to control them.
The shift moves AI “safety” from enforceable internal guardrails to publicly stated goals the company says it can change, as the Pentagon pressures it over contract access and red lines.
In practice, a company that branded itself around restraint is now signaling it will keep building more powerful systems without the former stop-trigger, even while resisting demands tied to weapons control and mass domestic surveillance.
Reality Check
When a Cabinet secretary ties a $200 million contract and an “effective government blacklist” to rolling back safeguards, we are watching coercive procurement power get repurposed as policy enforcement outside democratic process, and it will not stop with AI companies. The conduct described risks crossing into criminal territory if it functions as an unlawful quid pro quo or extortion under color of official right—potentially implicating the Hobbs Act (18 U.S.C. § 1951) and federal bribery/gratuities frameworks (18 U.S.C. § 201)—and even short of that, it is a textbook abuse-of-office pressure tactic. Threatening to invoke the Defense Production Act and label a firm a “supply chain risk” to force weakened safeguards is the kind of weaponized discretion that chills dissent and weakens our ability to demand limits on mass domestic surveillance. If this becomes routine, our rights become negotiable terms in federal contracts rather than binding constraints on government power.
Legal Summary
Level 2 exposure: the article describes high-level executive pressure tying a major Pentagon contract and potential blacklisting to Anthropic’s rollback of safeguards, raising serious investigative concerns about improper influence and politicized procurement. However, the facts presented do not allege personal enrichment or a thing-of-value to an official, so the classic structural quid-pro-quo corruption elements are not established in the article.
Legal Analysis
<h3>18 U.S.C. § 201 — Bribery / Illegal Gratuities (public officials)</h3><ul><li>The article describes a threatened loss of a $200M Pentagon contract and potential blacklisting unless Anthropic rolls back safeguards; that is coercive leverage tied to contracting, but there is no alleged thing-of-value paid to an official or personal enrichment.</li><li>Facts provided show pressure and conditional government action, not a described corrupt exchange for private benefit; quid-pro-quo elements (payment/gift to the official) are not alleged.</li></ul><h3>18 U.S.C. § 872 — Extortion by officers or employees of the United States (color of official right)</h3><ul><li>Defense Secretary allegedly issued an ultimatum: change safety policy by a deadline or risk loss of contract/blacklist; this reflects use of official power to compel private conduct.</li><li>However, the article does not allege the official sought or obtained “money or other thing of value” for himself; the demanded concession appears policy/compliance-oriented, leaving a key statutory element uncertain on these facts.</li></ul><h3>41 U.S.C. ch. 21 (Procurement Integrity Act) / 48 C.F.R. (FAR) — Improper influence in federal procurement (procedural integrity)</h3><ul><li>Threatening to cut off a contractor/offeror unless it alters internal safeguards could be viewed as politicized or retaliatory pressure affecting procurement decisions.</li><li>The article does not describe misuse of source-selection information, kickbacks, or specific procurement communications that satisfy statutory/procurement-integrity prohibitions; exposure is primarily an investigative red flag regarding improper influence.</li></ul><h3>18 U.S.C. § 1505 — Obstruction of proceedings (administrative/agency)</h3><ul><li>The alleged threat to “blacklist” and to invoke the Defense Production Act to designate a “supply chain risk” could implicate misuse of administrative levers.</li><li>No specific pending “proceeding” and corrupt intent to obstruct it are described; facts currently align more with pressure tactics than chargeable obstruction.</li></ul><b>Conclusion:</b> The reported conduct most strongly reflects procedural/administrative coercion and potential procurement politicization (investigative red flags) rather than a money-access-benefit transactional corruption pattern or clearly satisfied criminal bribery elements on the facts provided.</p>
Media
Detail
<p>Anthropic announced in a Tuesday blog post that it is changing its two-year-old Responsible Scaling Policy, replacing self-imposed guardrails with a more flexible, nonbinding framework it says “can and will change.” The company said shortcomings in the prior policy could hinder its ability to compete in a rapidly growing AI market and that industry peers did not adopt similar constraints.</p><p>The new approach includes a “Frontier Safety Roadmap” describing the company’s guidelines and safeguards, which it characterized as “public goals” it will “openly grade” rather than “hard commitments.” Anthropic also said it will separate its own safety plans from its recommendations for the broader AI industry.</p><p>The change followed a meeting with Defense Secretary Pete Hegseth and came a day after Hegseth reportedly set a Friday deadline for Anthropic CEO Dario Amodei to roll back safeguards or risk losing a $200 million Pentagon contract and being placed on what was described as an effective government blacklist. A source said Anthropic is unwilling to drop concerns about AI-controlled weapons and mass domestic surveillance of American citizens, and Hegseth is reported to be considering invoking the Defense Production Act and designating the company a supply chain risk.</p>