Anthropic's large language model, Mythos, has been deemed too dangerous to release by cybersecurity experts and is exploiting vulnerabilities in every major operating system and internet browser. The company claims that Mythos identifies over 99 percent of the thousands of vulnerabilities that it has identified, many unnoticed for decades. Anthropic is launching an initiative to safely deploy these AI capabilities and enhance cybersecurity. However, concerns have been raised among AI researchers about the company's claims. The full extent of regulatory authority over Mythos remains unclear, with some arguing that the Defense Production Act allows the president to force companies to prioritise government contracts and manufacturing important defense products, but it doesn't give the government complete authority over the model.