A report said Anthropic will release a public version of its new AI model Claude Mythos on June 10 local time.

Sources.news, a media outlet run by journalist Alex Heath (알렉스 히스), reported that the public version of Mythos will have a substantial level of guardrails. It will have more limited cyber security functions than the version accessible to Project Glasswing partners.

But Sources.news said its ability to handle complex, multi-step tasks over long periods has improved significantly.

Earlier, Anthropic announced it would expand the Glasswing project in early June to 150 organisations in more than 15 countries.

The Glasswing project is an initiative to help users find vulnerabilities and prepare response measures using Anthropic's AI model.

Anthropic provided access to the Claude Mythos preview, Mythos Preview, to 50 initial partners including the U.S. government in early April. Newly participating organisations cover the power, water, healthcare, telecommunications and hardware sectors, which were not heavily represented in the initial group. "The commonality across each partner is that catastrophic consequences could result if their codebase is attacked," Anthropic said. "A major attack could affect more than 100 million people," it said.