unknown

Security researchers discovered vulnerabilities in OpenAI's Atlas browser and other AI agents that allow websites to serve manipulated content to AI crawlers while showing different information to human visitors. By detecting AI user-agent headers, malicious actors could inject false information into LLM responses, manipulate job recruiting processes, or launch smear campaigns. Additional vulnerabilities include inadequate anti-phishing protections and insecure OAuth token storage. Meanwhile, research shows US businesses lag globally in AI governance, with only 17.5% having governance programs in place despite aggressive AI adoption.

Exclusive: OpenAI’s Atlas browser — and others — can be tricked by manipulated web content

The official daily.dev AI community. Led by thought leaders and AI experts. Stay tuned for insights, news and discussions. 