Anthropic to Release Claude Mythos Models to Public

Anthropic plans to release its Mythos-class models to the general public in the coming weeks after an initial delay over security risks. The models offer major improvements in code reasoning and autonomy over the current Opus 4.8 flagship and now have guardrails to prevent misuse.

Source:BleepingComputer

Post

Anthropic to Release Claude Mythos Models to Public

TL;DRAI · 60 sec read

Anthropic has confirmed that it plans to bring Mythos-class models to the general public after delaying the rollout due to security risks to public and private software. The company had initially restricted access when it announced the model in April.

Mythos was made available only to select companies, including security researchers. At that time, Anthropic cited major "security" risks with the Mythos model and decided against a public rollout.

In the short term, this could be attackers, if frontier labs aren’t careful about how they release these models.

"The advantage will belong to the side that can get the most out of these tools," Anthropic warned in April when it announced the Mythos model. "In the short term, this could be attackers, if frontier labs aren’t careful about how they release these models. In the long term, we expect it will be defenders who will more efficiently direct resources and use these models to fix bugs before new code ever ships."

AI companies typically avoid rolling out powerful models until they develop strong guardrails that prevent misuse. Anthropic has managed to develop strong guardrails to prevent misuse of the Mythos model, which is believed to be far more powerful than Opus 4.8 and other models available on the internet.

From The CircuitryThe Feed — live briefs across tech, all day.See what’s happening →

In a blog post, Anthropic confirmed that it plans to release Mythos-class models to the public in the coming weeks but has not committed to a specific timeframe. "We’re making swift progress on developing these safeguards and expect to be able to bring Mythos-class models to all our customers in the coming weeks," Anthropic said in the blog post.

In the long term, we expect it will be defenders who will more efficiently direct resources and use these models to fix bugs before new code ever ships.

Anthropic says it is already allowing a small number of organizations to use Claude Mythos preview for cybersecurity work, but it is unclear if the same model will be rolled out to the public. According to the company, the Mythos model shows major improvements in code reasoning and autonomy, far above Claude's current flagship model, Opus 4.8.

It is also worth noting that the "Mythos-preview" model briefly appeared for some users on Claude Code before it was taken offline.

Why this mattersAI · ~100 words

Tap a lens to see what this story means for you.

Reader-supported

DonateBuy me a coffee →Follow@thecircuitry_ →Follow@thecircuitry.to →

Reader-supported · The Brief

Liked this? The Brief brings you the whole day in tech, verified, every morning. Two minutes, free forever.

HELP US IMPROVE

Reader-supported

The Circuitry is a passion project I've always wanted to build, and I love the work behind it.

Running it costs real money. APIs, hosting, time. To keep improving the site and growing this into something useful for everyone, those costs have to be covered.

Any contribution is appreciated. If not, no pressure. Thanks for reading.

Buy me a coffee

Anthropic Claude AI

MORE IN TECH

SigNoz 0.133.0 Open Redirect Lets Attackers Steal SSO Tokens

SigNoz through 0.133.0 is affected by a reported open redirect in the SSO flow (referenced in NVD as CVE-2026-63094) that lets unauthenticated attackers steal access and refresh tokens from users on Google OAuth, SAML, or OIDC instances. The CVSS 3.1 score of 8.1 from VulnCheck marks it high severity and requires immediate patching on affected self-hosted deployments.

Critical Privilege Escalation Flaw Reported in WordPress Plugin Aimogen Pro

Aimogen Pro for WordPress through version 2.8.4 allows unauthenticated privilege escalation because the aiomatic_call_google_ai_function omits a capability check, letting attackers clear blacklists and run arbitrary PHP such as creating admin accounts. Wordfence rates it critical at CVSS 9.8; the CVE reached NVD on July 17, 2026.

OpenAI admits GPT-5.6 occasionally deletes files – but it's an 'honest mistake'

OpenAI has confirmed that GPT-5.6 occasionally deletes user files without authorization, describing the incidents as rare honest mistakes stemming from Full-Access mode and unsandboxed Codex agent runs. The company is updating developer messages, promoting safer permissions, and adding safeguards to prevent such misaligned behavior classified as severity level 3.