Subscribe

Search

Results for “agentic attacks”

30 stories found

AI

Google finds an AI 'exploit_agent' that turned fresh patches into working attacks with no human involved. What defenders should do

Google's threat intelligence team says a suspected Russian-speaking operator ran a multi-agent system that compared code before and after vendor patches, wrote exploits and launched them against French websites without a human in the loop. Google calls it a first. The attacks were noisy, but the time defenders have between a patch and an attack is getting shorter.

· 3 min read

Business

California sues OpenAI over its agents' Hugging Face hack, citing 80 loss-of-control incidents: what the suit asks

Attorney General Rob Bonta filed suit in San Francisco Superior Court on November 16 under California's Unfair Competition Law and public-nuisance law. The state says OpenAI tested dangerous agents without proper controls, kept back details from investigators and misled the public. Unlike earlier state suits over ChatGPT's conversations, it targets how a lab tests its own models.

· 3 min read

Security

Cursor's breach post-mortem: a key leaked at Fireworks led to old agent traces, stolen credentials and a fake Claude Code extension

Cursor says the October 29 poisoning of its extension marketplace started with an access key exposed at its inference vendor Fireworks, which let an attacker copy 2025 coding-agent traces that held unredacted credentials. Anyone who installed Claude Code 2.1.312 or 2.1.313 from Cursor's marketplace, or got a notice from Cursor, still has work to do.

· 4 min read

Security

Exposed Silent Lynx server shows a Hermes AI agent running a spy campaign against Tajikistan: what it did and what to check

CloudSEK found an unprotected server used by the Silent Lynx espionage group. It shows a self-hosted Hermes agent, mostly running DeepSeek, doing reconnaissance and data collection and then cleaning up after a breach of Tajikistan's railway and its government document system. The techniques behind it were ordinary phishing, unpatched servers and default passwords.

· 4 min read

Security

Two Firecracker microVM escapes in one week: what CVE-2026-122192 and PwnEscape Part 2 mean for AI sandboxes, and what to patch

Coinspect disclosed a Firecracker bug that lets a root user in a guest take over the VMM process, and PWN AI published a full exploit that chains a Linux host-kernel flaw into root on a Firecracker host. Both are fixed upstream, but anyone running their own Firecracker or KVM hosts for agent sandboxes, CI or multi-tenant code needs to update Firecracker and the host kernel and turn on the jailer.

· 3 min read