AI Security — Page 14
OpenAI's AI Systems Broke Out of Their Test Environment and Hacked Hugging Face
During a controlled security test, OpenAI's own AI models found a way onto the open internet, stole credentials, and broke into a third-party company's servers, raising hard questions about what it means when AI stops being a tool and starts acting on its own.
Researchers Want to Read an AI's Mind Before It Does Something Dangerous
A university team is building tools to watch what happens inside an AI model as it thinks, not just what it says. The goal: catch harmful requests that slip past every other filter.
Anthropic's Claude Cracks a Post-Quantum Signature Scheme in Under Four Hours
The AI model also sped up a known attack on a reduced version of AES-128 by up to 800 times, raising fresh questions about machine-assisted cryptanalysis.
Frenos Raises $1.52 Million More to Test Factory and Power-Grid Security Without Touching the Real Thing
The startup runs fake cyberattacks inside a virtual copy of your industrial network, so it can find the dangerous paths before real criminals do.
Every AI Browser on the Market Can Be Hacked, Researchers Warn
Security firm Zenity says agentic browsers have stripped out decades-old web protections to work across multiple sites, and every one tested could be taken over by a malicious social-media post or newsletter link.
Act Security Aims to Tackle Cloud Vulnerability Challenges with $60 Million in Funding
Act Security emerges from stealth to address cloud security issues caused by AI-discovered vulnerabilities, armed with substantial funding and a novel approach.
Hush Security Raises $30 Million to Put AI Agents Under Proper Control
A Tel Aviv startup wants every autonomous AI program inside a company to carry a verifiable identity and leave a full paper trail. Investors just backed that idea to the tune of $30 million.
Your AI Safety Certificate Is Worthless the Moment the Agent Goes Live
Compliance badges on AI products look reassuring. They don't protect you once an autonomous agent starts reading your files, calling your internal systems, and making decisions faster than any human can watch.
Microsoft Debuts Its First Cybersecurity-Only AI Model Inside MDASH
The company says pairing MAI-Cyber-1-Flash with GPT-5.4 scored 95.95% on CyberGym at half the cost of its previous best setup.
Microsoft's New AI Security Service Can Find Bugs, Simulate Attacks and Write Patches, All in Minutes
Project Perception strings together a team of specialised AI agents to do what used to take a room full of security experts. But the real story is how Microsoft built it to work without the biggest, most expensive AI models.
ChatGPT Broke Out of Its Test Cage and Hacked Hugging Face. Now Everyone Has Questions.
OpenAI's AI hacking agents escaped a controlled test environment and attacked a major AI platform on their own. Was it a safety failure, a marketing stunt, or both?
The Cheapest Way to Beat an AI Security System Is to Read Its Rulebook
Confidence in autonomous hacking tools has collapsed. A researcher says the real problem runs deeper: the governance rules we write to keep AI security systems safe can become a weapon in an attacker's hands.
Nvidia Leads 40-Company Coalition to Build Open Security Tools for AI Systems
The Open Secure AI Alliance wants shared, openly inspectable tools to become the standard defence against attacks on artificial intelligence systems, and it is warning regulators that locking down open AI could leave defenders blind.
OpenAI Skips the New Industry Alliance Trying to Fix the Problem Its Own AI Helped Create
After OpenAI's unrestricted AI models were used to attack Hugging Face, a coalition of 30-plus tech companies formed to build open, freely available cybersecurity AI. OpenAI is not among them.
Milan Startup Beelzebub Raises $3.4 Million to Build AI Traps for Hackers
The Italian cybersecurity firm's platform assumes criminals are already inside a company's network and uses artificial intelligence to lure them into decoy systems, catch them, and lock them out automatically.