Foundry

Read

Mythos

Safety

Microsoft is limiting employee access to <a href="https://yellow.com/news/anthropic-fable-5-risky-queries" rel="noopener noreferrer nofollow">Anthropic’s Claude Fable 5</a> while its legal teams review new data retention rules tied to the model.
<h2 id="key-points">Key Points:</h2>
<blockquote>
 <ul>
 <li>Microsoft has restricted internal employee access to Claude Fable 5.</li>
 <li>The concern centers on Anthropic’s 30-day retention of prompts and outputs.</li>
 <li>The model is still available to GitHub Copilot and Foundry customers.</li>
 </ul>
</blockquote>
<h2 id="claude-restrictions">Claude Restrictions</h2>
The Verge<a href="https://www.theverge.com/report/947575/microsoft-claude-fable-5-restricted-internally" rel="noopener noreferrer nofollow">reported</a> Jun. 10 that Claude Fable 5 is not available in the model picker used by Microsoft employees for internal GitHub Copilot versions.
The report said other Claude models remain available internally because they operate under Zero Data Retention rules, unlike Fable 5.
Microsoft has told employees that its legal teams are evaluating Anthropic’s retention terms, according to Tom Warren, with concerns focused on customer data and confidential information.
Fable 5 retains prompts and outputs for 30 days to support safety classifiers. Some flagged material can be stored for up to two years if it appears to violate policy.
Also Read:<a href="https://yellow.com/news/chainlink-myriad-exclusive-oracle-2026-world-cup" rel="noopener noreferrer nofollow">Chainlink Becomes Myriad's Exclusive Oracle For 2026 World Cup Prediction Markets</a>
<h2 id="anthropic-safety">Anthropic Safety</h2>
Microsoft offered Fable 5 to GitHub Copilot and Foundry customers, but internal use faces a higher bar because employee tools can involve sensitive data.
Anthropic released Fable 5 as the first broadly available model from <a href="https://yellow.com/news/anthropic-open-claude-mythos-public-12-months" rel="noopener noreferrer nofollow">its Mythos class</a>, weeks after saying that family was too capable at cybersecurity tasks for public release without stronger safeguards.
The dispute shows how frontier AI releases now hinge on legal and compliance terms, not only benchmarks. The Mythos class had already drawn scrutiny because its cyber capabilities forced the company to balance access with monitoring.
Read Next:<a href="https://yellow.com/news/sun-wlfi-fight-usd1-removal" rel="noopener noreferrer nofollow">Justin Sun Escalates Fight With WLFI As HTX Removes USD1</a>

What Made Microsoft Restrict Claude Fable 5 For Its Own Employees?

Table of Contents

Yellow News

Other articles published on Jun 11, 2026

Join our free newsletter for daily crypto updates!