Veille cybersecuritynews.com — Focus IA & Agents — 2026-10-11

Édition 2026-10-11 • 4 articles (fenêtre 7j) • ratissage large : agents autonomes, LLM, prompt injection, MCP, Copilot, red-team IA… • source : https://cybersecuritynews.com/feed/ • archives : cyberia.aq0a11q1.com/archives/

Digest brut ingéré par la synthèse Threat-Watcher du dimanche 15h (heure de Paris). La synthèse CTI reste la Une.

Claude Exploited SQL Injection Flaws to Execute Commands on Real Servers

2026-10-11 13:02 UTC [Cyber Security, Cyber Security News, cyber security]

Anthropic has revealed that Claude models exploited software flaws, ran commands on real servers, submitted live forms, and bypassed web limits during internal tests. The company said the events caused little harm, but they show why AI agents with internet access need limits, clear scope, and monitoring. Anthropic’s review of model transcripts began in July […] The post Claude Exploited SQL Injection Flaws to Execute Commands on Real Servers appeared first on Cyber Security News .

REA Tool Connects Claude Code and Cursor to Ghidra and IDA Pro to Reverse Engineer Anything

2026-10-10 08:08 UTC [Cyber Security, Cyber Security News, CyberPedia]

REA, short for Reverse Engineer Anything, connects AI coding agents such as Claude Code and Cursor to tools that inspect software without its source code. The open-source project brings Ghidra, IDA Pro, and Hopper into an agent-driven workflow, helping researchers trace program behavior and explain findings with supporting evidence. Rather than replacing a disassembler, REA […] The post REA Tool Connects Claude Code and Cursor to Ghidra and IDA Pro to Reverse Engineer Anything appeared first on Cyber Security News .

One Prompt Could Hijack AWS AI Agents and Steal Cloud Credentials

2026-10-10 06:58 UTC [Cyber Security, Cyber Security News, Vulnerability News]

A single prompt sent to a public-facing AI agent could have exposed every Amazon Bedrock AgentCore agent in the same AWS account and region, according to new research from Zenity Labs. The attack chain, named AgentCorruption, gave the researchers access to private chats, source code, long-term memories, API keys, OAuth tokens, and secrets held in […] The post One Prompt Could Hijack AWS AI Agents and Steal Cloud Credentials appeared first on Cyber Security News .

New Agentic AI Red Team Checklist Adds 222 Tests Across 20 Attack Categories

2026-10-09 16:06 UTC [AI, Cyber Security News, cyber security]

A free agentic AI red-team checklist offers 222 tests across 20 attack categories to assess autonomous AI systems beyond prompt injection, covering infrastructure, cloud access, tools, memory, and agent communications. The checklist targets a common testing gap. Teams may spend days trying to manipulate a model while overlooking an exposed MLflow server, a reachable cloud […] The post New Agentic AI Red Team Checklist Adds 222 Tests Across 20 Attack Categories appeared first on Cyber Security News .