Tag
44 articles
Cal has abandoned its open-source model for its flagship program due to security concerns from AI-powered threats, describing the situation as 'like handing out the blueprint to a bank vault.'
Security researcher Aonan Guan exploited prompt injection flaws in AI agents from Anthropic, Google, and Microsoft, stealing API keys. All three companies paid bug bounties but did not issue public advisories.
OpenAI expands its Trusted Access for Cyber program with GPT-5.4-Cyber, enhancing cybersecurity capabilities for vetted defenders while strengthening safeguards.
A software developer claims to have reverse-engineered Google DeepMind's SynthID watermarking system, raising questions about the security of AI-generated content protection.
Anthropic has decided to keep its latest AI model private after discovering thousands of cybersecurity vulnerabilities across major operating systems and web browsers. The company chose to quietly share the findings with organizations responsible for internet infrastructure instead of releasing the model.
London-based Trent AI has raised $13 million to develop agentic security solutions for autonomous AI systems. The startup, led by former Amazon and Cambridge experts, aims to protect multi-agent AI environments as they become more prevalent.
Microsoft releases an open-source toolkit to enhance runtime security for AI agents, addressing growing concerns about autonomous language models executing code and bypassing traditional controls.
Apple, Google, Microsoft, and 9 other tech giants join forces under Project Glasswing to use Anthropic's Mythos AI model for finding critical software vulnerabilities before adversaries do.
Anthropic's new AI model has identified security vulnerabilities across major operating systems and web browsers, marking a significant advancement in automated cybersecurity. The model is part of Project Glasswing, a collaboration between tech giants including Nvidia, Google, Microsoft, and Apple.
Asylon and Thrive Logic partner to bring physical AI into enterprise perimeter security, combining autonomous robotics with AI analytics for real-time threat detection.
Hackers are distributing leaked Claude AI source code with added malware, while the FBI warns of a national security risk from a recent hack of its wiretap tools.
Learn to analyze and defend against AI agentic tools like OpenClaw that can exploit system vulnerabilities for unauthorized access. This tutorial covers network monitoring, vulnerability scanning, and access control strategies.
