Cybersecurity Companies’ Stocks Fall as Anthropic Tests Powerful New Model

Anthropic is testing a new, highly capable AI model called Mythos that outperforms prior models in code reasoning and cybersecurity benchmarks and can autonomously surface previously unknown vulnerabilities; the company has limited access to vetted customers and warned of dual-use risks. Markets reacted negatively—cybersecurity equities dropped—amid concerns that such AI could enable rapid, large-scale exploitation if safeguards fail or adversaries replicate the capability, with Anthropic noting past interest from a state-sponsored actor in earlier models.