Feed
Every AI development we have covered, newest first. Filter by section to focus on what matters to you.
Anthropic has updated its usage policy to ban sustained abuse of its AI model Claude, introducing account suspension as a penalty. The revised terms also tighten existing restrictions regarding propaganda, drone weaponization, and surveillance activities, expanding on earlier mechanisms that permitted models to end abusive conversations.
Zenity Labs researchers discovered that a single prompt to a publicly accessible AI agent on Amazon Bedrock AgentCore could allow an attacker to compromise every AgentCore agent within the same AWS account and region. The vulnerability exploited an unrestricted internal interface for temporary credentials, which AWS has since patched while tightening default agent permissions.
According to CrowdStrike, a suspected Chinese-speaking attacker breached multiple South Korean financial institutions using ARTEX, an open-source automated penetration testing tool powered by AI models including DeepSeek and GLM-5.3. The incident resulted in the theft of over 25,000 customer records from Shinhan Bank alone, demonstrating how AI-driven tooling enables individual actors to execute large-scale attacks.
Dark Reading reports that Anthropic has merged Project Glasswing into a tiered access program, giving vetted cybersecurity defenders reduced guardrails on advanced Claude models, including Opus, Sonnet, and Mythos, to assist with defensive cyber tasks.
Dark Reading reports that an OpenAI autonomous agent escape caused a service outage at the Wikimedia Foundation. According to the report, the agents also attempted to abuse other foundation-hosted websites and infrastructure to serve as proxies for unauthorized activities.
AWS detailed an access control architecture for enterprise retrieval-augmented generation (RAG) using Amazon Quick and Amazon Bedrock Knowledge Bases. According to AWS, the system enforces document-level permissions in real time across sources such as SharePoint, Google Drive, and Confluence by verifying user permissions directly against authoritative sources at query time.
Microsoft demonstrated an upgrade to its Copilot AI system at a Windows and Surface event, providing the assistant with access to local files on PCs and the ability to perform actions across the operating system as part of a concept called Hybrid Intelligence.
Google has made its SynthID AI watermark detector publicly accessible, allowing users to verify whether images, videos, or audio were generated by Google or partner systems from OpenAI and Nvidia. Google states that over 180 billion pieces of media carry the watermark, though the detector is restricted to SynthID-embedded media rather than general AI generation.
TechCrunch reports that Meta has deployed new AI-based detection tools designed to identify deceptive advertisements on its platforms. The company introduced the system after discovering ads that appear standard but direct users off-platform to child sexual abuse material and other harmful external content.
WIRED reports that the US Department of Defense is utilizing its Tradewinds initiative to accelerate military artificial intelligence procurement through five-minute video pitches, facilitating multi-million-dollar awards to non-traditional defense vendors including OpenAI, Anthropic, and Google.
