Tag: AI Security
29 articles tagged AI Security
Litigant hid instructions in filings to sway an AI reader
A self-represented claimant buried invisible prompts in court documents telling any model reading them to agree with him. Court staff spotted the white space.
UK researcher backdoors an open AI model for under £75
Manchester Metropolitan's Katie Paxton-Fear implanted a backdoor in an open-weight model in about an hour for under £75 ($100), using ten training examples.
UK AISI finds GPT-5.5 matches Claude Mythos on vulnerability-finding
The UK AI Security Institute has found OpenAI's generally-available GPT-5.5 performs comparably to Claude Mythos on security vulnerability-finding evaluations.
McKinsey's internal AI chatbot breached via basic SQL injection
Security firm CodeWall used an autonomous AI penetration tool to access McKinsey's Lilli chatbot, exposing 46.5 million chat messages and system configurations.
European Parliament Disables AI Features on Lawmakers' Devices
The European Parliament has switched off AI tools on official devices after its IT department could not guarantee where data was being sent.
Researchers Find 40,000 Exposed OpenClaw Instances on Public Internet
SecurityScorecard found over 40,000 exposed OpenClaw instances, with 63% of deployments vulnerable and nearly 13,000 exploitable via remote code execution.
UK Government Partners With Microsoft on Deepfake Detection Framework
The Home Office announces a new deepfake detection evaluation framework developed with Microsoft, aiming to set industry standards as deepfake volumes surge.
Researchers Warn Open-Source AI Models Vulnerable to Criminal Misuse
Joint SentinelOne and Censys research reveals thousands of open-source LLM deployments operating without guardrails, creating security risks for spam, phishing and disinformation.
GenAI and AI Agents Create New Class of Insider Risk, Research Warns
Proofpoint report reveals 40% of organisations concerned about data loss via GenAI tools, with AI agents operating as privileged superusers creating critical threats.