InfoAI Security
Three charged in $129 million fraud using AI-cloned executive voices
Prosecutors say the scheme impersonated executives and lawyers; one Missouri hospital system lost $4.8 million.
DeepfakeBEC
Trust Infinitech
AI Security
3 storiesMediumAI Security
Patched bug let attackers bypass ChatGPT's Lockdown Mode
Lockdown Mode, added in June to limit data theft through prompt injection, could be sidestepped before OpenAI fixed the flaw.
Prompt injectionLLM
The Hacker News
HighAI Security
GitBook says AI-agent attack reached infrastructure credentials
Integration tokens and API tokens may be exposed. GitBook revoked all personal API tokens and its MCP server needs re-authentication.
AI agentsAPI tokensMCP
explainx tracker