A cybersecurity researcher who once rolled his eyes at AI doomsday talk, then started issuing warnings, is now the guy that Trump's White House is flying in to calm its nerves. The Wall Street Journal traces how Anthropic's Nicholas Carlini—long known in the field as a "professional skeptic" of AI risk—tested the company's powerful Mythos model earlier this year and came away alarmed. "I don't think we should release Mythos yet," he wrote to colleagues at the time.
Using prompts now nicknamed the "Carlini Loop," he had the AI churn through mountains of code and turn up hundreds of software bugs, including one in the Linux operating system and a serious flaw in the Ghost publishing platform. Within weeks, hackers were exploiting the Ghost bug in the wild. That proof-of-concept helped trigger what some in cybersecurity circles are calling "Bugmageddon": the realization that AI can now find and help weaponize vulnerabilities at scale. The clash that followed—between Anthropic and a now more hawkish Trump administration—has produced emergency bans on the firm's newest models, tense Washington negotiations, and bigger questions about whether systems like Mythos 5, as well as Anthropic's supposedly more secure Fable 5, can ever be safely released.
Still, last week, Anthropic sent Carlini to Washington "to soothe the government's nerves" and to explain that "even though there was no such thing as guaranteed safety in AI, it was better for the world to release Fable than to keep it under wraps," per the Journal. Crypto Briefing notes "the fracturing of what was once a relatively unified safety community," adding that that community has broken off into at least three distinct segments: "those who still want hard restrictions, those who favor controlled access models like Anthropic's consortium approach, and those who believe openness itself is a safety strategy because it enables broader scrutiny."
,
uBlock 
,
and
Ghostery 
)
