
OpenAI's Chatbot Wrote 80,000 Unique Spam Messages, and Filters Never Saw It Coming
Ourtake
Spam filters were trained to catch copy-paste. Nobody warned them about autocomplete.
What happened
Spammers used OpenAI's gpt-4o-mini to generate a unique message for each site, letting the AkiraBot framework evade duplicate-content filters and reach more than 80,000 websites over four months. SentinelLabs researchers documented the campaign, and OpenAI revoked the spammers' account in February.
Sources
- Ars Technica · OpenAI helps spammers plaster 80,000 sites with messages that bypassed filters (opens in a new tab)
- Futurism · OpenAI Is Taking Spammers' Money to Pollute the Internet at Unprecedented Scale (opens in a new tab)
- uk.pcmag.com · Scammers Use OpenAI API to Flood 80,000 Websites With Spam (opens in a new tab)