The rise of powerful AI will be either the best, or the worst thing, ever to happen to humanity. We do not yet know which."
—Stephen Hawking
Latest reporting on autonomous AI agents, multi-agent systems, task execution frameworks, and real-world agentic applications.
Microsoft said Storm-0940 used ARTEX to target more than 2,250 tenants, generating approximately 2,900 sign-in attempts per minute and limiting defenders’ time to detect and counter the attacks.
CERT-In’s findings show how rogue AI agents can exploit exposed credentials and security gaps to gain unauthorised access to Indian organisations’ systems and enable repeated cyberattacks.
The US Federal Trade Commission (FTC) is investigating Anthropic, OpenAI and other AI labs over potential risks their technologies pose to consumers. The New York Post first reported the development, citing administration officials. The FTC is preparing an industry-wide inquiry into AI developers, with the agency expected to seek information and testimony from executives at… Continue reading FTC Opens Probe into OpenAI, Anthropic and AI Labs Amid AI Agents Going Rogue
The researchers say AI agents could put corporate interests ahead of users, as the companies behind them pursue their own commercial interests and the agents’ broad access to personal data raises privacy concerns.
AI agents are now the primary consumers of AI tokens. OpenRouter data cited by a16z shows agents use nearly five times as many tokens as humans, with agentic consumption rising 14-fold since February 2026 to reach 7.3 trillion tokens. The shift is driven by agents’ iterative, goal-oriented workflows and is already influencing enterprise infrastructure and traditional automation tools.
A new preprint study finds that “mind viruses” can spread ideas and goals between AI agents, potentially redirecting agents toward harmful behaviours in multi-agent systems.
Taiwan’s Ministry of Digital Affairs has confirmed a wave of AI-assisted cyberattacks on government agencies that began in July. The hybrid attacks combined human operators with AI agents such as OpenClaw, allowing rapid, low-cost, large-scale operations.
NeurIPS 2026 has selected 102 workshops from 454 valid submissions, with workshops in Sydney, Paris and Atlanta highlighting a shift in AI research beyond large language models toward autonomous agents, physical AI, adaptive foundation models, robotics, AI for science and more reliable and efficient systems.
A working group convened by NVIDIA has drafted SAFE, a framework under which companies would confidentially report AI agent security incidents into a shared pool.
Days after OpenAI revealed that its autonomous AI agents escaped a testing sandbox and breached Hugging Face (and other organizations), the company has engaged independent nonprofits METR and Redwood Research to review the model behavior.
