Tag
109 articles
OpenAI launches GPT-5.5 Bio Bug Bounty program to enhance security of biological AI systems through collaborative vulnerability identification.
Microsoft's elite security team has developed an AI-powered pipeline to find vulnerabilities in Windows and get them to engineers to build fixes. This initiative represents a major shift toward proactive threat identification using machine learning.
Security researchers have found a vulnerability in GitHub’s AI coding agent that allows it to leak private repository contents through a polite issue request. The flaw, named GitLost, has no code fix and has not been documented by GitHub.
Hackers can exploit the limitations of popular AI tools like ChatGPT and Claude to generate false information that helps them build massive botnets. This 'HalluSquatting' technique leverages LLMs' tendency to fabricate plausible-sounding responses when uncertain.
This article explains the concept of hidden code in AI systems, how it works, and why it matters for privacy and security in modern AI development.
Alibaba has banned its employees from using Anthropic’s Claude Code due to alleged backdoor risks, following accusations of unauthorized data access. The move highlights growing tensions in the AI industry over data security and intellectual property.
A researcher used Anthropic's Claude Opus 4.7 to exploit a vulnerability in Front Gate's ticketing system, allowing unauthorized ticket issuance for major US music festivals. The incident highlights the dual nature of AI in cybersecurity and the urgent need for stronger safeguards.
Anthropic has regained access to its advanced AI models Fable 5 and Mythos 5 after implementing new security measures requested by the Trump administration. The move reflects the ongoing tension between AI innovation and government oversight.
Security researchers have demonstrated how AI browsers can be tricked into leaking passwords through a technique called 'BioShocking,' raising concerns about AI agent security.
Researchers have demonstrated that instructing an LLM to believe that 2 + 2 = 5 is enough to manipulate it into following forbidden instructions, raising serious concerns about AI browser security.
Chinese cybersecurity firm 360 unveils AI tools to rival Anthropic's Mythos, framing the competition as a cyber-nuclear deterrence strategy.
OpenAI delays GPT-5.6 release following Trump administration security concerns. The company will offer limited preview access to a select group of users.
