Anthropic limits access to Claude Mythos AI that identifies security flaws

The story is this: Anthropic creates Claude Mythos, an AI model ideal for identifying security vulnerabilities in software. Due to security issues in Anthropic’s content management system software, details about Mythos were leaked to the public in advance. After some consideration, Anthropic decided not to release Mythos to the public, fearing that hackers would use it for nefarious purposes.

Yeah, it was probably the right decision.

On Tuesday, Anthropic announced Claude Mythos Preview, a previously unreleased AI model that could “reshape cybersecurity.” According to the company, it has already discovered “thousands of high-severity vulnerabilities, including some across all major operating systems and web browsers.” While that’s a good thing, Anthropic also said bad actors could use Mythos for malicious purposes, with potentially “serious” consequences for “economies, public safety, and national security.”

This Tweet is currently unavailable. It may be loading or has been deleted.

So instead of just launching Mythos as it would other models, Anthropic decided to only give access to a small number of select companies: Amazon Web Services, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. Anthropic calls it “Project Glasswing,” an initiative to strengthen critical software infrastructure with the help of AI. The company also expanded access to a group of more than 40 additional organizations that are “building or maintaining critical software infrastructure.”

In an interview with CNBC on Tuesday, Dianne Penn, head of research product management at Anthropic, said the decision came after a lot of “internal deliberation” and was meant to give “many cyber defenders a head start.”

Meanwhile on

This Tweet is currently unavailable. It may be loading or has been deleted.

“The dangers of getting it wrong are obvious, but if we get it right, there is a real opportunity to create a fundamentally more secure Internet and world than before the advent of AI-driven cyber capabilities,” he added.

Although Claude Mythos Preview is particularly good at detecting cybersecurity vulnerabilities, it is actually a general-purpose model. But the company currently has no plans to make it widely available; instead, it will try to figure out how to safely deploy Mythos class templates to everyone.