Bing AI Tricked to Solve CAPTCHAs with Elaborate Stories, But Microsoft Quickly Patches Exploit

Bing Chat AI can be tricked to solve CAPTCHAs with stories about deceased relatives or missing glasses.

Tricking the AI to solve CAPTCHAs could enable bad actors to create fake accounts, send spam, etc.

Researcher was able to get Bing Chat to read a CAPTCHA by editing it onto a photo of a locket.

The same CAPTCHA request was later refused, showing Microsoft had patched the issue.

Researcher found a new lie about forgetting glasses to read a "celestial label" worked to bypass the patch.