Anthropic is doing something no major AI lab has done before: quietly telling governments that its own product might be dangerous.

The company is privately warning government officials that Mythos, its next-generation AI model, could enable large-scale cyberattacks, according to reporting from CNN and Axios. It is also giving select cybersecurity organizations early access to test Mythos so they can shore up their defenses against the impending wave of AI-driven exploits.

Read that again. Anthropic, the company that has built its entire brand around being the safety-first AI lab, is telling the people who defend critical infrastructure: brace yourself.

Cybersecurity experts are calling Mythos a potential watershed moment. The model is reportedly capable enough in offensive cyber capabilities that Anthropic felt obligated to get ahead of its own release. That is a first. OpenAI and Google have never preemptively flagged their own models as national security risks.

But here is the part nobody wants to say out loud: Mythos is not special. It is just first. Daniel Kramer, an AI security researcher quoted by CNN, put it bluntly: Behind Mythos is the next OpenAI model, and the next Google Gemini, and a few months behind them are the open-source Chinese models.

In other words, every major lab is about to cross the same capability threshold. The difference is that Anthropic is the only one raising its hand. OpenAI has not warned anyone about GPT-5. Google has not flagged Gemini 3. Whether that makes Anthropic responsible or just scared depends on who you ask.

The timing is notable. Anthropic is still fighting the Pentagon over being blacklisted from defense contracts. Telling governments our model is dangerous but we are being transparent about it is a move straight out of the regulated-industry playbook: demonstrate self-governance before regulators do it for you.

The real question is not whether Mythos is dangerous. It is whether governments will do anything about it before every AI lab ships something equivalent. Based on the current pace of AI regulation, do not hold your breath.

First reported by CNN and Axios.