anthropic-mythos-broad-release-2026

fields

statementREQ	Anthropic's eventual goal is to enable users to safely deploy Mythos-class models at scale, contingent on developing cybersecurity safeguards; Mythos Preview will NOT be made generally available in the near term.	www.anthropic.com	2026-04-30
actorREQ	Anthropic	www.anthropic.com	2026-04-30
topic	model safety / access policy	www.anthropic.com	2026-04-30
promised_at	2026-04-13	www.anthropic.com	2026-04-30
target_date
shipped	false	www.anthropic.com	2026-04-30
shipped_at
evidence_url	https://www.anthropic.com/glasswing	www.anthropic.com	2026-04-30

history · 6 fields · 6 revisions

statement1 revision

Anthropic's eventual goal is to enable users to safely deploy Mythos-class models at scale, contingent on developing cybersecurity safeguards; Mythos Preview will NOT be made generally available in the near term. current www.anthropic.com · 2026-04-30

We do not plan to make Claude Mythos Preview generally available, but our eventual goal is to enable our users to safely deploy Mythos-class models at scale... We need to make progress in developing cybersecurity (and other) safeguards that detect and block the model’s most dangerous outputs.

actor1 revision

Anthropic current www.anthropic.com · 2026-04-30

We do not plan to make Claude Mythos Preview generally available, but our eventual goal is to enable our users to safely deploy Mythos-class models at scale... We need to make progress in developing cybersecurity (and other) safeguards that detect and block the model’s most dangerous outputs.

topic1 revision

model safety / access policy current www.anthropic.com · 2026-04-30

We do not plan to make Claude Mythos Preview generally available, but our eventual goal is to enable our users to safely deploy Mythos-class models at scale... We need to make progress in developing cybersecurity (and other) safeguards that detect and block the model’s most dangerous outputs.

promised_at1 revision

2026-04-13 current www.anthropic.com · 2026-04-30

We do not plan to make Claude Mythos Preview generally available, but our eventual goal is to enable our users to safely deploy Mythos-class models at scale... We need to make progress in developing cybersecurity (and other) safeguards that detect and block the model’s most dangerous outputs.

shipped1 revision

false current www.anthropic.com · 2026-04-30

We do not plan to make Claude Mythos Preview generally available, but our eventual goal is to enable our users to safely deploy Mythos-class models at scale... We need to make progress in developing cybersecurity (and other) safeguards that detect and block the model’s most dangerous outputs.

evidence_url1 revision

https://www.anthropic.com/glasswing current www.anthropic.com · 2026-04-30

We do not plan to make Claude Mythos Preview generally available, but our eventual goal is to enable our users to safely deploy Mythos-class models at scale... We need to make progress in developing cybersecurity (and other) safeguards that detect and block the model’s most dangerous outputs.