Anthropic is finally willing to give you Mythos, but it comes with some safeguard strings attached.
On Tuesday, Anthropic said it was releasing a “Mythos-class” model called Claude Fable 5, but with safeguards so significant that it gave the model a different name.
“Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage,” Anthropic said in its announcement.
The safeguards, which mean some queries on certain topics will instead be ported to Anthropic’s Opus 4.8 model, are initially tuned to be more conservative.
Anthropic is also continuing Project Glasswing, which it first announced in April, that gave early access to the first version of Mythos, Claude Mythos Preview, to a limited number of tech, cybersecurity, and financial companies. The AI company said “a small group of cyberdefenders and infrastructure providers” will get access to Claude Mythos 5, which is the same model as Fable 5 “but with the safeguards lifted in some areas.”
“It has the strongest cybersecurity capabilities of any model in the world,” the company said of Mythos 5. “Soon, we intend to expand access to Mythos 5 through a broader trusted access program.”
Even with constraints, Anthropic said Fable 5 “exceed those of any model we’ve ever made generally available.”
“It is state-of-the-art on nearly all tested benchmarks of AI capability, showing exceptional performance in software engineering, knowledge work, vision, scientific research, and many other areas,” the company said. “The longer and more complex the task, the larger Fable 5’s lead over our other models.”
Anthropic said Stripe found that Fable 5 compressed months of engineering into days. For financial tasks, the company said IMC found that Fable 5 “aced their trading-analysis evaluations nearly across the board, including factual lookup, conceptual reasoning, root-cause analysis, and expected-value analysis.”
Like previous Anthropic models, Fable fits within the theme of naming AI models after forms of composition. The AI startup said Fable was chosen because it shares similarities with Mythos.
“Fable is from the Latin fabula, ‘that which is told,’ akin to the Greek mythos,” the company said.
The release comes at a critical time for Anthropic, which recently announced that it had confidentially filed a draft S-1, officially kicking off the march toward a blockbuster IPO.
OpenAI, Anthropic’s chief rival, announced on Monday that it too had confidentially filed its S-1. SpaceX, which includes Elon Musk’s xAI, is set to go public on Friday when it begins trading on the Nasdaq.
Anthropic’s initial announcement of Mythos spooked financial markets and governments worldwide, raising concerns that AI models had advanced to the point where they could now uncover major vulnerabilities in software and cybersecurity.
President Donald Trump recently signed an executive order that allows AI companies to voluntarily give the federal government access to frontier AI models up to 30 days before their release. The order explicitly prohibits the government from imposing a mandatory review on AI companies.
Read the full article here
