AI Companies Cite Nuclear Arms Control While Undermining Journalism That Made It Possible

Leaders at Anthropic are calling for AI safety measures modeled on Cold War nuclear verification regimes—even as their industry undermines the journalism that made those historic agreements possible.

Marina Favaro and Jack Clark of Anthropic recently advocated for building a "brake pedal" into AI development, citing the now-expired Intermediate-Range Nuclear Forces Treaty as precedent. Clark told CNN that Cold War rivals "found ways to stabilize aspects of the nuclear arms race" under tense conditions.

The comparison arrives months after the New START arms control pact expired in 2026, following the U.S. withdrawal from the INF Treaty in 2019. But the analogy overlooks a critical element: Nuclear arms control succeeded not only through diplomatic bargaining but through public pressure informed by trusted, resource-intensive journalism.

The journalism crisis

New York Times publisher A.G. Sulzberger told the World Association of News Publishers in June that tech companies "strip-mine news websites without permission or compensation," repackaging content while diverting audiences and revenue from original sources. The Times has filed lawsuits against multiple AI corporations for copyright violations in training data and ongoing content use.

Google's recent "biggest upgrade to our Search box in over 25 years" prioritizes AI-generated answers over links to original reporting. Early users report that articles and research studies receive minimal visibility in results.

The six major AI companies have reached a combined valuation of $11 trillion, according to figures cited by Sulzberger. Meanwhile, news organizations face continued revenue decline after losing advertising to social media and search engines.

Historical precedent

Journalism has repeatedly proved essential to nuclear accountability. After the 1945 atomic bombings, reporters defied U.S. government denials to document radiation effects at Hiroshima and Nagasaki through on-the-ground investigation. Conversely, journalists' reliance on official sources without adequate scrutiny contributed to failures in reporting on Iraq's alleged weapons of mass destruction before the 2003 war.

Governments have consistently exploited public unfamiliarity with nuclear weapons to shape narratives, making independent verification through journalism crucial.

Why it matters

The AI industry faces growing calls for safety oversight comparable to nuclear regulation, yet its business model actively degrades the investigative capacity needed to hold powerful institutions accountable. Research by the BBC and European Broadcast Union found that 45% of AI-generated news responses contained at least one significant issue, with 81% showing some problems when evaluated for accuracy and context. As fewer journalists remain to conduct expensive original reporting on complex subjects—from nuclear policy to AI development itself—the public loses access to the verified information necessary for democratic oversight of existential risks.

Public preference remains

A Reuters Institute study found only 12% of respondents comfortable with entirely AI-produced news, compared to 62% for human-authored journalism. Americans mirror these global averages. However, preference alone won't preserve quality reporting without regulatory intervention.

The European Broadcast Union advocates for publisher control over AI training data use, with clear attribution and prominent links to original material when authorized. Sulzberger has emphasized direct audience relationships as the clearest path forward, though legal frameworks for AI regulation and content compensation will likely prove necessary across jurisdictions.

These details were first reported by The Bulletin of the Atomic Scientists.