Anthropic suspended access to two of its AI models for nearly three weeks in June 2026, then restored them the day after U.S. export controls were lifted. The company has now embedded a notice about the episode inside the system prompt for Claude Opus 5 to prevent misinformation.
A Brief Suspension and Restoration
Claude Fable 5 and Claude Mythos 5 first released on June 9, 2026. Just three days later, on June 12, Anthropic suspended access to both models. The move was made to comply with U.S. Department of Commerce export controls.
The controls remained in place for over two weeks. The Department lifted them on June 30, 2026. Anthropic restored access the very next day, on July 1, 2026.
The company published a statement about the events at https://www.anthropic.com/news/fable-mythos-access. That statement now serves as the reference point for the models' own explanations.
Why the Notice Exists
The suspension happened after Claude's training-data cutoff. That means the models have no direct memory of the events from their training. Without intervention, they could give outdated or incorrect answers.
To address this, Anthropic added a notice to the system prompt for Claude Opus 5. The notice tells the model what happened and directs it to the official statement. This ensures Claude doesn't provide incorrect answers about the export controls situation.
Simon Willison, a blogger and collector of quotations, posted the system prompt quotation on 9th August 2026. The quotation has drawn attention because it shows how a company handles a factual gap in a model's knowledge.
How Claude Responds
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
When asked about the suspension, Claude confirms the events accurately and matter-of-factly. It does not deny the suspension happened. It treats export controls like any other current political topic.
Claude gives a fair, accurate account rather than sharing personal opinions. It points to the linked statement for further information. The model also checks for newer information when it can search. If no newer information is found, it suggests checking Anthropic's site.
The notice is designed to handle a specific factual situation. The export controls are a political topic, and Claude is instructed to be neutral and factual. The system prompt ensures the model stays within those bounds.
A Model for Handling Post-Cutoff Events
The episode offers a glimpse into how AI companies manage events that occur after a model's training cutoff. Rather than letting the model guess, Anthropic chose to embed a corrective notice directly into the system prompt.
The approach has a clear benefit. Users asking about the suspension get a consistent, accurate answer. They also get a link to the official statement, which reduces the chance of confusion.
Things may have developed since the notice was written. That is why Claude suggests checking Anthropic's site for the latest information. The model's responses remain conditional on what it can find through search.
The quotation posted by Willison is not an original news report. It is a direct look at the system prompt itself. Still, it reveals how a major AI company handles a real-world disruption without letting its models spread falsehoods.
The timeline is straightforward. Release on June 9, suspension on June 12, controls lifted on June 30, restoration on July 1. The notice in Claude Opus 5's system prompt ties it all together.