will anthropic make mythos publicly available?
Anthropic has said it plans to open up Mythos—its restricted flagship AI model, from which Fable is made with additional safety guardrails—“more broadly in the future,” though access is currently limited to roughly 150 organizations across more than 15 countries. Mythos was released last month to a “small number” of partners for use in defensive cybersecurity, Anthropic said, indicating its latest version could be used to “find and exploit software vulnerabilities more effectively than any other model—and all but the most skilled human security experts.”
Read more Viral German Soccer Fan ‘Freddy’ Resumes Documenting U.S. Trip After Deactivating Account
anthropic’s rivalry with openai
Competition between Anthropic and OpenAI has become increasingly public in recent months. Anthropic CEO Dario Amodei, who spent nearly five years at OpenAI before leaving to co-found Anthropic in 2021, appeared to say last month he departed the company over his distrust of OpenAI CEO Sam Altman: “At the end of the day, why argue with someone when you don’t have the same vision and you don’t trust them?” That came months after Altman criticized Anthropic for running an ad during the Super Bowl, accusing the rival firm of being “so clearly dishonest” and calling it “authoritarian,” arguing OpenAI had no intention of serving ads in the way Anthropic portrayed.
Read more Elon Musk’s Fortune Falls Below $900 Billion As SpaceX Slide Worsens
key background
Anthropic debuted Fable in early June as a restricted variant of its more powerful Mythos model. The company noted Fable included safety features that blocked users from inquiring about cybersecurity, as well as biology and chemistry needed to create bioweapons. The Commerce Department forced Anthropic to pull the model days later, however, and the AI firm disclosed the government “became aware” of a report from researchers at Amazon who were able to bypass Fable’s safety mechanisms. Researchers were able to get Fable to “demonstrate” how a software security vulnerability could be exploited, according to Anthropic, which noted the technique used by the researchers was now “blocked in over 99% of cases.” Anthropic has been at odds with the Trump administration for months, after the Pentagon designated the firm a supply chain risk in March following a spat between officials and Anthropic, which refused to allow the government unrestricted access to its models.
Read more Northern Lights Update: 9 States Could See The Aurora Borealis Tonight
further reading
ForbesAnthropic’s Fable And Mythos Models Are Back—Here’s What The Government Was Concerned About In The First PlaceBy Zachary Folk