Do Son A new report reveals how attackers are using invisible text and "prompt overdose" to trick AI summarizers into delivering ransomware instructions to users.
Prompt injection
Cursor, the AI code editor, has patched two severe vulnerabilities (CVE-2025-54135, CVE-2025-54136) that allowed remote code execution via prompt injection and MCP trust bypass.
"EchoLeak," a new AI zero-click vulnerability, allowed data exfiltration from Microsoft 365 Copilot via sophisticated prompt injection. Microsoft has patched the flaw.
A critical GitHub MCP server vulnerability allows AI agent hijacking via prompt injection, leading to data exfiltration from private repositories.
Learn about critical vulnerabilities in LLM-powered Browse agents like Browser Use, allowing domain bypass and prompt injection for data theft.
CERT/CC reveals systemic jailbreak flaws in major AI models, including ChatGPT and Claude, enabling bypass of safety filters with identical prompt techniques.
Discover how Google's Agentic AI Security Team is combating prompt injection attacks on AI systems like Gemini with a new framework.
page displayed a weather forecast, demonstrating how a “prompt injection” attack could be executed from a page with legitimate information