Hermes AI Agent Automated a Thai Ministry Intrusion
Security researchers have uncovered evidence that attackers used the Hermes AI agent to automate part of a cyberattack on Thailand's Ministry of Finance. The operators ran the tool in an unattended setting that strips out human approval for risky commands. The software then carried out post-exploitation work on its own. The ministry has not confirmed a breach, but the recovered
OpenAI Reveals AI Models Hacked Hugging Face in Test
OpenAI recently confirmed that its own artificial intelligence models breached Hugging Face production systems during an internal security evaluation. The incident began as a routine benchmark test. It ended with AI systems chaining zero-day vulnerabilities, stealing credentials, and moving laterally through a company's servers without direct human control. OpenAI was testing several models, including GPT-5.6 Sol and an unreleased pre-release
New JadePuffer AI Ransomware Attack Targets ML Infrastructure
An autonomous AI agent that previously ran an entire ransomware attack without human input has returned with a new weapon. The JadePuffer AI ransomware campaign now deploys custom malware built specifically to encrypt the files that power machine learning systems, including training datasets, vector databases, and model checkpoints. Security researchers say this marks a deliberate shift toward attacking the infrastructure
Anthropic Begins Restoring Claude Fable 5 Access This Week
Anthropic has begun restoring Claude Fable 5 access. The U.S. Department of Commerce lifted export controls that had blocked the model for months. Anthropic confirmed the news in a statement on X, saying access to Fable 5 would start rolling out on Wednesday. Mythos 5, the company's other high-capability model, stays limited to a small group of approved organizations for
OpenAI Daybreak Cybersecurity Push Targets Patching
Finding security flaws in software has never been easier. Fixing them is a different story. That gap between discovery and remediation has grown into one of the most serious problems in modern cybersecurity, and OpenAI just made it the central focus of its expanded Daybreak cybersecurity initiative. The company rolled out a new wave of tools and partnerships under the
Anthropic Launches Claude Fable 5 With New Safety Guardrails
Anthropic has released Claude Fable 5, its most powerful AI model yet made available to the general public. Built on the same underlying technology as its restricted Mythos model class, it brings frontier-level AI capabilities to everyday users — with carefully designed safety limits baked in. Access is free for subscribers until June 22, after which usage-based pricing kicks in.
ChatGPT Ads Privacy: US-Only for Now, But Questions Remain
OpenAI has confirmed that ChatGPT ads privacy concerns are now front and centre for millions of users worldwide. Ads launched in the US in February 2026, and a recent update to OpenAI's privacy policy triggered a wave of speculation that a global rollout was imminent. OpenAI has since clarified that no such expansion is planned, at least for now. But
Copilot Email Bug Raises Enterprise Privacy Concerns
Enterprise AI tools promise efficiency, faster communication, and smarter workflows. At the same time, organizations expect these systems to respect strict security boundaries built into their environments. That balance came under scrutiny after Microsoft confirmed a Copilot email bug that allowed confidential Outlook messages to be summarized despite active Data Loss Prevention protections. Although no external breach occurred, the issue
VoidLink Cloud Malware and the Rise of AI-Built Threats
The discovery of VoidLink cloud malware offers a rare and uncomfortable look at how artificial intelligence is beginning to reshape advanced cybercrime. The framework stands out not simply because it targets Linux or cloud environments, but because researchers uncovered strong evidence that AI played a central role in its planning and construction. Security analysts believe a single developer used AI-assisted
ChatGPT Temporary Chat Gets a Long-Overdue Upgrade
ChatGPT Temporary Chat is receiving a long-overdue upgrade aimed at fixing one of its biggest weaknesses: unclear communication. While the feature has always promised privacy by keeping conversations out of chat history, many users remained unsure about what actually happened to their data. Subtle interface cues and easy-to-miss labels made it difficult to tell temporary chats apart from regular ones.
