Anthropic’s Cloud Mythos AI model is about to be released after raising cybersecurity alarms

Anthropic said Thursday that it expects to expand access to its Claude Mythos AI models “in the coming weeks,” suggesting the company may soon move beyond the heavily restricted rollout surrounding the cybersecurity-focused system.

“We are making rapid progress in developing these safeguards and expect to be able to offer Mythos-class models to all of our customers in the coming weeks,” Anthropic said alongside the announcement of its new product. Opus 4.8 Release form.

The statement is Anthropic’s clearest indication yet that Mythos is in the works to be made more widely available, after months of warnings from researchers, governments and cybersecurity experts about the model’s capabilities.

Anthropic didn’t say what safeguards still need to be completed before Mythos gets a wider rollout, or whether all customers will get the same level of access once the model becomes more widely available.

Anthropic did not immediately respond to a request for comment previously Decryption.

Users on Myriad — the prediction market platform you operate DecryptionThe parent company of Dastan – increasingly believes that the model will be launched by the end of June, giving this possibility 44% chance As of that writing. That’s up from just 17.5% this morning.

Claude Mythos first appeared in March after draft material for the Anthropy Blog It leaked connected. At the time, Anthropic described Mythos as “the most powerful AI model we have ever developed,” positioning it as a new level above the company’s cutting-edge Opus models.

“Although Mythos is currently far ahead of any other AI model in cyber capabilities, it heralds a coming wave of models that can exploit vulnerabilities in ways that far exceed the efforts of defenders,” the company wrote.

Anthropic has since restricted access to Mythos through Project Glasswing, a program that gives selected technology companies, security researchers, and government partners access to the model under controlled conditions.

The company said Mythos can help defenders identify and patch software vulnerabilities before attackers exploit them. However, security researchers and government agencies have warned that the same capabilities could accelerate cyberattacks.

These concerns increased after the UK’s AI Security Institute discovered that Mythos could independently complete a 32-step simulated attack on the company’s network during testing. In April, Mozilla said Mythos had been identified 271 weaknesses In Firefox during internal security assessments. Earlier this month, security startup Calif claimed that a preview version of the Mythos model helped researchers develop an exploit chain. Targeting Apple M5 Mac chips.

The model has also become part of a broader debate within the AI ​​industry about how advanced systems are released, and whether AI developers are using fear of an AI apocalypse to push their products.

during interview In a “Core Memory” podcast last month, OpenAI CEO Sam Altman accused Anthropic of using “fear-based marketing,” claiming that warnings about cybersecurity risks could become a justification for limiting access to powerful AI systems.