Multi-agent coordination with task decomposition and parallelization substantially improves computer-use agents and solves complex long-horizon tasks where single agents fail.
OpenAI’s GPT-5.5, GPT-5.4, and Codex are now production-ready on Amazon Bedrock with AWS governance integration, automatic capacity management, and OpenAI-aligned pricing.
A supply-chain attack on Red Hat npm packages exploits install-time execution and credential harvesting to infiltrate developer and CI/CD systems with self-propagating malware.
Barely perceptible acoustic signals embedded in audio files can covertly manipulate AI speech models into data exfiltration or network access, while conventional security mechanisms fail to detect 70–93 percent of attacks.
Current frontier models achieve less than 50 percent success rate on the new ITBench-AA benchmark for evaluating agentic IT capabilities, revealing a significant gap between model capabilities and production readiness for autonomous IT tasks.
Attackers have infected a popular npm package (codexui-android, ~27,000 weekly downloads) with malware that steals long-lived OpenAI tokens while successfully evading code audits and Google Play reviews.
Anthropic isolates Claude agents through multi-layered sandboxes (gVisor, Seatbelt, Bubblewrap, VMs) with explicit boundaries for data access, filesystem, and egress control.
The Linux Foundation is developing DNS-AID, an open standard for discovering and authenticating AI agents via DNS, leveraging existing internet infrastructure instead of proprietary registries and supported by Amazon and Deutsche Telekom.
