BitGo CEO Mike Belshe has rejected viral claims that Anthropic's Mythos model breached nearly all of the National Security Agency's classified systems, calling the story false as it spread across social media this weekend. According to reports from The Economist, Belshe answered one of the threads bluntly, stating "I'm calling BS on this." The rejection targets posts that recast the government shutdown of a three-day-old model as a real-world hack, with the fuller record being less dramatic than the viral narrative suggests.
The viral claim originated with Senator Mark Warner, vice chair of the Senate Intelligence Committee, who cited what the Economist reported as the NSA director's account. Warner said General Joshua Rudd described the tool as breaking into almost all of their classified systems in hours, not weeks, not weeks. However, as reported by The Economist, Warner raised the example while praising Anthropic, not condemning it, using it to argue for faster pre-release testing of frontier models. The detail missing from online reports is that this was an authorized red-team test on the agency's own networks, not an outside intrusion. The US government was already a Mythos partner, with Anthropic having deployed the model to government cyber defenders through Project Glasswing since April.
Zack Korman mocked how the claim moved from senator to journalist to social media unchecked, while analyst Kyle Chase noted the break-in was a test. According to The Economist, Shashank Joshi, the editor who published the quote, later cautioned it should not be read literally, stating it depended on Mythos working alongside other tools in particular conditions. Anthropic's own statement supports the test explanation, stating the flagged jailbreak simply asked the model to read a codebase and fix flaws, surfacing minor bugs that rival models like OpenAI's GPT-5.5 can also find. A separate jailbreak flagged by Amazon was identified as the real trigger for the shutdown.
The company disabled both models on June 12 to meet a US export-control directive, not because of any battlefield breach, as reported by The Economist. Anthropic objected to recalling a model used by hundreds of millions of people over one narrow flaw. Whether the test justified pulling the models is still contested, with AI researcher Pedro Domingos arguing the export controls were responsible given the model's powerful hacking capabilities. The company is now working to restore access and is drafting a shared risk framework with the White House. Anthropic itself calls Mythos the strongest cyber model in the world, yet it argues that recalling a tool over one flaw would freeze new releases across the industry.
