Anthropic says its Mythos AI tool is too risky to release to the general public because of its ability to identify software and computer system vulnerabilities. The company argues that if such a capability were widely available, it could be misused by attackers to steal data or disrupt critical infrastructure. According to multiple reports, Anthropic initially limited Mythos to a small set of carefully chosen partner organizations, rather than offering it broadly.
The outlets also report that Anthropic is increasing the number of organizations that can use Mythos. As of early June, Anthropic granted access to 150 additional organizations, bringing the total user group to roughly 200 organizations across 15 countries. The reports frame this as a deliberate approach: while Anthropic restricts public availability due to safety concerns, it expands access through controlled partnerships with selected organizations. The coverage does not indicate that Anthropic is changing its assessment that a public release would create unacceptable risk; instead, it highlights the contrast between limiting public deployment and widening authorized, vetted use through partners.
