Anthropic’s newest AI tool, Claude Mythos 5, has finally rolled out to the public after the company felt it was too powerful to release earlier in April.
The model follows Claude Fable 5, a private preview that stirred concerns among technology, finance and government leaders over its potential to hack or exploit systems. Anthropic said the public release would include "safeguards and user limitations," but admits the model’s capabilities exceed those of any previously released version.
In the testing period, a small group of about 150 organisations identified more than 10,000 security flaws, prompting the company to tighten access to "trusted cyber‑defenders and infrastructure providers" before opening it wider. Despite an ongoing lawsuit with the U.S. Department of Defense, federal agencies have been experimenting with Mythos, underscoring the tool’s high interest even among regulators.
Canadian Finance Minister François‑Philippe Champagne highlighted that the model’s mystery—"the unknown, unknown"—justifies heightened scrutiny, while U.S. government agencies are still evaluating its use for national security.
Anthropic’s valuation has now approached $1 trillion, suggesting investors value the company’s broader AI capabilities alongside its risk management strategy. The firm plans to "expand access to Mythos 5 through a broader trusted access program" in the near future.
Co‑founder Jack Clark warned on BBC Newsnight that the AI industry’s rapid growth needs a "brake pedal," or a way to slow deployment. The current public release marks a pivotal moment in balancing innovation with safety.
For more context, read the BBC articles on the May 4th preview and the April 25th announcement.Finance ministers debate risks and what’s happening with Narrative?
