Gates told Ezra Klein that cyberattack capabilities capable of disrupting bank accounts and the power grid already exist today, leaving AI in a period of extreme vulnerability. He cited Anthropic's Mythos model and Project Glasswing, arguing that open-source model risks come from humans rather than AI sentience, and called for a supervisory layer with absolutes on all models.
At the Qatar Economic Forum, Citigroup CEO Jane Fraser warned that increasingly powerful AI models could trigger cyberattacks at any time, saying companies are racing to shore up defenses and a 'tsunami' of patching work is coming. She also said the technology bottleneck has shifted from chips to energy, and called the release of Anthropic's Mythos model 'not a good day.'
Anthropic announced the opening of applications for the Life Sciences Verification Program (LSVP), which enables life science professionals—including first-time users of the Mythos model—to conduct biology-related work under new safeguards, balancing research needs with misuse risk. The beta program is open to academic labs, startups, pharma companies, and other teams, with future expansion to individual Pro and Max users.
Anthropic released the Claude5 series, including Fable5 for the public and Mythos5 focused on safety. Fable5 excels in software development, scientific analysis