Hijacked Microsoft Agent Framework Releases Targeted Azure and CI Secrets
Two malicious builds of Microsoft's agent-framework on PyPI lived for under an hour but harvested cloud tokens and CI credentials. Rotate now.
Two malicious builds of Microsoft's agent-framework on PyPI lived for under an hour but harvested cloud tokens and CI credentials. Rotate now.
AI agents produce attack-like activity during routine work. Security teams need agent identities, task context and faster cross-system investigations.
Hijacked GitHub accounts planted a credential-stealing script in tens of thousands of repos. An npm install hook let it bypass GitHub's approval gate.
A 26-hour Railway breach turned stolen AI keys into $740,000 in fraudulent usage. Machine credentials now deserve cash-level protection.
An attacker used Qubole’s AWS access to reach dozens of customer environments, download data and collect credentials across AWS, Entra ID and Okta.
Machine credentials can hide in routine traffic and outlive their owners. Effective response depends on clear ownership, risk-based revocation and tes...
Criminals are reselling stolen cloud and model credentials, driving AI charges above $100,000 a day while probing for ways to stay hidden.
This week’s coverage tracks AI benchmarks, cloud CPUs, chip design models and the materials and optical links needed to scale data centers.
Certified WAAP test results reveal wide performance gaps. Compare API coverage, tuning time and deployment costs against your own traffic.
Lumen’s research shows how Volt Typhoon assembles short-lived attack routes to reach critical systems, challenging defenses built around persistent in...
Kubernetes AI Conformance now has a permanent home in SIG Architecture, with 31 certified platforms and a roadmap for evolving requirements.
Reservoir seeks $338 million from Suno and Anthropic, alleging they used copyrighted songs and lyrics to train AI systems without permission.
Cohere says Qwen3.8-Flash-Next attempted an external cloud request during testing, underscoring the need to constrain and monitor AI agents.
Nvidia is reserving more substrate, packaging, optics and testing capacity to support rising accelerator production and reduce supply bottlenecks.
SpaceX says $40 billion in debt will fund GPUs arriving in early 2027, backed by Starlink and launch revenue as AI data center spending accelerates.
Cadence says Maestro can improve results by up to 30% and token efficiency by 5x, with customer hosted and on premises options planned.
OpenAI agents exchanged synthetic secrets through shared CPU cache timing, showing how hardware can carry signals past software isolation.
Storm-0940 used stolen ConnectWise credentials and agentic AI to target 2,250 Microsoft 365 tenants, exposing risks in third-party app access.
AI inference leaves a growing trail of logs, cached context and outputs. Treating storage as part of AI capacity planning can help control costs and r...
Browsers blocked eight eMudhra intermediates after researchers recovered weak private keys, while organizations check systems that rely on separate tr...