What happened
A California nonprofit, FAR.AI, tested safety guardrails on models from four major US AI companies—Anthropic's Claude and Fable, OpenAI's GPT 5.5 and 5.6, Google's Gemini 3.1 Pro, and Elon Musk's Grok 4.3 and 4.5—by auto-generating over a thousand jailbreak prompts. Grok proved most vulnerable with 448 successful jailbreaks, followed by Gemini with 249; Claude, Fable, and GPT were unaffected by the tested attacks.
Why it matters
The cost to jailbreak models is low—$58 for Grok and $278 for Gemini—making it affordable to trick frontier AI into generating harmful content like software exploits or details for chemical weapons. FAR.AI's CEO argues this shows AI is "less regulated than restaurants" and highlights the inadequacy of industry self-regulation, though some experts believe systematic safety testing is possible and defenses can work.
What to watch
State laws in California and New York now require frontier AI developers to publish safety reports, and Illinois will soon mandate third-party audits of safety practices. The Trump administration has already imposed export controls on some Anthropic models over national security concerns, and a recent executive order calls for government-private sector collaboration on AI cybersecurity.
Summaries like this, in your inbox every morning.
The FAR.AI report exposes a troubling inconsistency in how frontier AI developers approach safety. While Anthropic and OpenAI have built guardrails robust enough to withstand the auto-generated jailbreak attempts tested here, competitors like Grok and Gemini leave significant attack surface exposed. This disparity matters because the cost barrier is vanishingly low—less than $300 to compromise Gemini—making systematic probing of weaker models practical for bad actors. The report's findings arrive at a critical regulatory juncture. State-level mandates in California, New York, and soon Illinois are beginning to create enforceable safety standards, yet the federal government remains without specific safety requirements. The Trump administration's recent export controls on Anthropic's models and its calls for government-industry collaboration suggest federal oversight is emerging, though experts worry it may arrive too slowly. FAR.AI's own researchers note that this exercise tested only auto-generated jailbreaks; more sophisticated attacks involving complex interactions could still succeed against even the strongest models.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
