The new AI model that could steal your life savings

Anthropic’s AI model, Claude Mythos, is very powerful. SO powerful it can find software vulnerabilities that might let it, I don’t know, steal your bank login information. Anthropic is holding back the model from a wider release for now. Today on the show, how worried should we really be about Mythos? And are its capabilities actually unique? <br /><br />The Indicator has a weekly newsletter! Sign up now: <a href="http://npr.org/newsletter/indicator" target="_blank">npr.org/indicatornewsletter</a> <br /><br /><strong>Related episodes:  </strong><br /><a href="https://www.npr.org/2025/10/06/nx-s1-5561881/fighting-ai-with-ai" target="_blank">Fighting AI with AI</a> <br /><a href="https://www.npr.org/2025/10/09/nx-s1-5562031/how-ai-might-mess-with-financial-markets" target="_blank">How AI might mess with financial models</a> <br /><br /><em>For sponsor-free episodes of The Indicator from Planet Money, subscribe to Planet Money+ via Apple Podcasts or at </em><a href="http://plus.npr.org/" target="_blank"><em>plus.npr.org</em></a><em>.</em> <em>Fact-checking by </em><a href="https://www.npr.org/people/1268825622/sierra-juarez" target="_blank"><em>Sierra Juarez</em></a><em>. </em><em>Music by </em><a href="https://dropelectric.bandcamp.com/" target="_blank"><em>Drop Electric</em></a><em>. Find us: </em><a href="https://www.tiktok.com/@planetmoney" target="_blank"><em>TikTok</em></a>, <a href="https://www.instagram.com/planetmoney/" target="_blank"><em>Instagram</em></a>, <a href="https://www.facebook.com/planetmoney" target="_blank"><em>Facebook</em></a>, <a href="http://npr.org/indicatornewsletter" target="_blank"><em>Newsletter</em></a><em>. </em> <br /><br /><br />See <a href="https://pcm.adswizz.com">pcm.adswizz.com</a> for information about our collection and use of personal data for sponsorship and to manage your podcast sponsorship preferences.<br /><br /><a href="https://www.npr.org/about-npr/179878450/privacy-policy">NPR Privacy Policy</a>