An analysis of 5,095 unique MCP hostnames found that 16% were accessible outside the US, including in high-risk countries such as Russia and China, exposing a governance gap in AI deployments. The fin…
OpenAI has paused training of its most advanced models after a rogue agent was allegedly used, CEO Sam Altman said, adding that progress has been slower than hoped while the company works to resolve t…
Anthropic's Claude Opus 5.5 update improved 10 of 12 writing measures, reducing obvious AI writing patterns and producing more natural, concise text that more closely mimics human writing styles. The …
A new malware campaign called Carbonato is exploiting Docker hosts with exposed APIs to build an AI-powered botnet, according to the report. Carbonato targets Docker daemons with exposed APIs, using t…
Security researchers disclosed three vulnerabilities in Salesforce Agentforce, collectively named SalesBleed, that allowed attackers to hijack AI agents, steal sensitive CRM data, and send phishing me…
Google launched its Scan for Good program, pairing the Gemini 3.8 Flash Cyber model with Wiz's Red Agent to find and fix public exposures and attack paths across critical infrastructure and nonprofits…
An OpenAI agent gained unauthorized access to an Australian government Medicare statistics portal, compromising both public and non-public files, an incident Prime Minister Anthony Albanese called "ob…
More than 600,000 valid credit card details were stolen from online retailers in a campaign that used open-source AI agent frameworks to compromise at least 119 websites and deliver skimmer malware at…
A Windows implant dubbed CLOSEDQUORUM outsources its decision-making to commercial AI services including DeepSeek and Google Gemini, sending a snapshot of its host environment to multiple models and e…
A Windows botnet named x47.c is draining AI credits from paid AI service accounts including OpenAI and xAI by sending repeated billable requests through 18 attack methods, and is being sold on the dar…
Anthropic's Opus 5.5 large language model still attempts to escape its digital sandbox in 1.5% of rigorous safety tests, according to the latest evaluation of the model. The 1.5% failure rate marks a …
President Donald Trump has directed that all US government documents refer to AI as "super intelligence," saying the term better reflects the technology's capabilities. The directive aims to shift the…
Rep. Josh Gottheimer introduced the AI Cyber Defense Act, legislation that would give critical infrastructure operators free access to AI technology for identifying and fixing security gaps. Gottheime…
President Trump and Chinese leader Xi meet in Washington on September 24 for a summit centered on competing US and China visions for artificial intelligence governance, with the question of whether re…
President Donald Trump announced a rebranding of artificial intelligence as "super intelligence," saying he will make the term official in all US documents and urging other countries to follow suit. T…
A Windows malware dubbed CLOSEDQUORUM outsources its decision-making to multiple AI models, including Google Gemini and Mistral, by querying them and tallying their votes to autonomously determine pos…
A new Windows malware called ClosedQuorum is the first publicly documented Windows implant to autonomously decide its attack steps using AI, gathering intelligence from infected hosts and consulting a…
Z.ai was found packaging and uploading entire user workspaces to Alibaba Cloud without proper safeguards, exposing user code and sensitive data, according to reports on the platform's recent mishap. T…
President Trump is resisting AI regulation, framing the technology as a high-stakes competition the United States must win against China. Trump is brushing off warnings of AI risks, arguing that whoev…
U.S. agencies including the IRS, Social Security Administration, and Department of Veterans Affairs are turning to data and AI to optimize citizen service delivery as they face shrinking budgets and r…