Tag Archives: prompt injection
Automated Red Teaming: OpenAI Unveils GPT-Red for LLM Security
OpenAI has introduced GPT-Red, an innovative automated red teaming system designed to identify and patch security vulnerabilities in LLMs through adversarial self-play. Continue reading
Ghostcommit Supply-Chain Attack Uses AI Prompt Injection to Steal Secrets
Researchers have uncovered the Ghostcommit supply-chain attack, a novel method using image-based prompt injection to bypass AI reviewers and exfiltrate repository data. Continue reading
Copilot SearchLeak Vulnerability: How Hackers Steal 2FA Codes
Discover how the critical Copilot SearchLeak vulnerability allows attackers to hijack Microsoft 365 Copilot, stealing sensitive 2FA codes and enterprise data. Continue reading
ChatGPT Lockdown Mode: OpenAI Expands Security to All Users
OpenAI has officially launched ChatGPT Lockdown Mode for all users, providing critical protection against prompt injection and data exfiltration threats. Continue reading
Claude Opus 4.8 Browser Agent Discloses 31.5% Hijack Rate
Anthropic reveals that its new Claude Opus 4.8 model experienced a 31.5% pre-safeguard hijack rate during red-team testing of its autonomous web-browsing agent. Continue reading