2026-03-27 🧭 Daily News

Mythos Leak, Pentagon Injunction & October IPO

🧭 Claude Mythos — Anthropic's Next Frontier Model Leaks from Unsecured Data Store

A data-handling error at Anthropic exposed the existence of a previously unannounced model called Claude Mythos. An unsecured internal data store surfaced a draft blog post describing Mythos as a "step change in capabilities" — more powerful than any model Anthropic has publicly released. The accidental disclosure, reported by Fortune on March 26, prompted Anthropic to confirm the model's existence while declining to provide a public release date.

What the leak revealed

Capabilities: The draft post highlighted exceptional performance in cybersecurity research, software programming, and academic reasoning — areas where Mythos reportedly outpaces the current Opus 4.6 generation.
Access: Mythos will initially be available only to select research and enterprise partners. No public rollout is planned in the near term.
Positioning: The "step change" language mirrors how Anthropic framed the jump from Claude 3 to Claude 4 — suggesting Mythos sits in a higher capability tier entirely.

Developer note

If you're building on the Anthropic API, keep an eye on platform.claude.com for early-access programme announcements. Mythos access will likely follow the same gated preview process used for Opus 4.6 extended thinking.

🧭 Federal Judge Grants Preliminary Injunction — Pentagon Ban on Anthropic Blocked

US District Judge Rita Lin issued a preliminary injunction on March 26 temporarily halting the Department of Defense's supply-chain risk designation against Anthropic. Judge Lin called the designation "classic First Amendment retaliation," finding that the government appeared to have targeted Anthropic specifically because it published safety commitments — including a refusal to allow Claude to be used for autonomous weapons targeting or surveillance of US citizens. The ruling is a significant procedural win for Anthropic as the broader lawsuit proceeds.

Key findings from Judge Lin's ruling

The DoD designation was issued after Anthropic published its AI safety constitution and usage policies — a sequence the court found suggestive of retaliatory motive.
Anthropic's published policy documents constitute protected speech under the First Amendment; the government had not demonstrated sufficient national-security justification to override that protection.
The injunction is preliminary and subject to revision as discovery proceeds. A full trial date has not yet been set.
Commercial enterprise customers using Claude through Azure, Google Cloud, and Amazon Bedrock are unaffected by the DoD designation and may continue normal operations.

What this means for enterprise procurement

Organisations that paused Claude evaluations due to the supply-chain risk designation can resume them. Anthropic's three major cloud partners (Microsoft, Google, Amazon) each publicly confirmed that their commercial offerings are outside the scope of the DoD order — and that position is now bolstered by the court's preliminary ruling.

🧭 Anthropic Weighs October 2026 IPO — Racing OpenAI to the Public Markets

Bloomberg reported on March 27 that Anthropic is actively exploring a public offering as early as October 2026, according to people familiar with the matter. The company has held preliminary conversations with investment banks and is monitoring market conditions. No formal decision has been made, and the timeline could shift — but the October window would place Anthropic ahead of rival OpenAI, which has also been linked to a potential 2026–2027 listing. These reports should be treated as indicative of intent, not a confirmed plan.

Context and caveats

Valuation baseline: Anthropic's most recent secondary-market transactions have been cited at approximately $380 billion (per analyst estimates reported by TechCrunch and Fortune). An IPO pricing would establish a market-determined figure — which may differ significantly from private-round estimates.
Revenue trajectory: Sacra estimated annualised revenue at $14 billion as of February 2026, a figure Anthropic has not confirmed publicly.
Pentagon uncertainty: The ongoing lawsuit with the DoD introduces legal uncertainty that could influence IPO timing decisions.
Lock-up for developers: An IPO would not change Anthropic's API pricing or model access in the near term; operator agreements and usage policies remain under Anthropic's control regardless of ownership structure.