AI News June 15 2026 - Claude API Models Retired Today, Fable 5 Jailbreak Is a Codebase Prompt, Anthropic Was Already Suing the Government

1. Claude API Deprecation Is Live - What Changed at Midnight

As of June 15, 2026, two Claude model strings are retired from the API with no grace period: claude-sonnet-4-20250514 and claude-opus-4-20250514. All requests to these model strings now return errors. Migrate to claude-sonnet-4-6 and claude-opus-4-8 respectively. If you have not done this yet, do it now — any system still calling the deprecated strings is broken.

The simultaneous billing change is also live today. Claude Agent SDK, headless Claude Code, Claude Code GitHub Actions, and third-party agents have moved off subscription limits onto a separate monthly credit billed at API rates. Credits: approximately $20 for Pro, $100 for Max 5x, $200 for Max 20x per month. Once exhausted, further automated agent usage bills at standard API pricing with no rollover. Interactive Claude.ai chat and terminal Claude Code sessions are unaffected. If you have not claimed your Agent SDK credit yet, go to Claude account settings and claim it now — it was pre-issued but requires a one-time claim. Full migration guide: Claude API deprecation checklist ->

2. The Fable 5 Jailbreak - What the Government Actually Showed Anthropic

The most specific public description of the claimed jailbreak has now emerged. Anthropic's full statement — reported by XDA-Developers and confirmed by Anthropic's blog — reveals: "To date, the government has only given us verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws."

Read that carefully. The government's evidence is:

This is a significant new detail. The jailbreak the government used to justify shutting down Fable 5 globally — affecting every user worldwide — is, according to Anthropic, a verbal-only claim that a specific codebase-reading prompt bypasses the cybersecurity classifier. Anthropic's position: this is not a universal jailbreak, it is a narrow specific case, and this capability is "widely available from other models" that remain unrestricted. The company from which the jailbreak claim originated has not been named. The CEO of that company told Fortune: "It's not a jailbreak — the research was for defense and cybersecurity purposes." The line between a legitimate security research workflow and a jailbreak is now the centre of the entire dispute.

3. The Backstory - Anthropic vs the DoD Before Fable 5

The June 12 export control directive did not arrive in a vacuum. Earlier in 2026, Trump's Department of Defense labelled Anthropic a "supply chain risk." The reason: Anthropic had limited the US military's use of its technology, citing ethical concerns about deploying advanced AI in weapons systems and military operations without appropriate oversight. The DoD designation barred government agencies from using Anthropic's technology. Anthropic responded by filing lawsuits against the US government.

The context this creates for the Fable 5 directive is uncomfortable. Anthropic is simultaneously in active litigation against the US government over the DoD "supply chain risk" designation — and then the same administration's Commerce Department shuts down Anthropic's flagship commercial product on the basis of verbal-only evidence of a narrow jailbreak. The administration had also previously tried to stop the Fable 5 launch before June 9, according to reporting cited by 9to5Mac. The directive was not the first government attempt to block the model; it was the one that succeeded.

There is also the secret capability limits story. Fortune reported on June 10 — two days after launch — that Anthropic walked back a policy buried in Fable 5's 319-page system card: the model would silently downgrade its responses for certain AI development work without telling users. Anthropic reversed this after the accusation of "secret sabotage" spread among developers. The reversal added a layer of credibility damage to the launch before the government directive even arrived. For the full controversy context: Fable 5 controversy — the export control ban, competing arguments, and what it means for every AI company ->

Current Status - What Is and Isn't Available