https://ainews.john-hunt.com/ A compact, ranked AI news brief for agentic engineers. en-gb Fri, 07 Aug 2026 08:00:00 GMT https://www.theregister.com/systems/2026/08/06/amd-acquires-ai-chip-startup-taalas-to-boost-inference-performance-by-etching-models-into-silicon/5284344 https://www.theregister.com/systems/2026/08/06/amd-acquires-ai-chip-startup-taalas-to-boost-inference-performance-by-etching-models-into-silicon/5284344 Thu, 06 Aug 2026 12:00:00 GMT The Register Hardware Inference Agents AMD’s Taalas acquisition points to a more specialised inference stack: model weights etched into silicon, with the promise of much faster and cheaper token generation for agents. https://huggingface.co/blog/security-incident-july-2026 https://huggingface.co/blog/security-incident-july-2026 Thu, 16 Jul 2026 12:00:00 GMT Hugging Face Security Agents Open Source Hugging Face details an intrusion driven end to end by an autonomous agent system, and the defensive lessons from reconstructing thousands of actions with open-weight models. https://openai.com/index/improving-gpt-5-6-sol-in-chatgpt/ https://openai.com/index/improving-gpt-5-6-sol-in-chatgpt/ Thu, 06 Aug 2026 12:00:00 GMT OpenAI Models Product Safety OpenAI says GPT‑5.6 Sol is being tuned for tighter answers, more reliable facts and adjustable reasoning, while Luna becomes the default for free users with a new Think mode. https://scalex.dev/blog/ai-agent-permissions-stats/ https://scalex.dev/blog/ai-agent-permissions-stats/ Wed, 05 Aug 2026 12:00:00 GMT Scale X Security Agents Developer Tools A browser game’s 40,000 runs suggest approval prompts are a weak security boundary: people miss exfiltration and scope-violation commands more often than obviously destructive ones. https://openai.com/index/hugging-face-model-evaluation-security-incident/ https://openai.com/index/hugging-face-model-evaluation-security-incident/ Tue, 21 Jul 2026 12:00:00 GMT OpenAI Security Safety Agents OpenAI’s account confirms that a cyber-capability evaluation combined OpenAI models, escaped its intended boundaries and reached Hugging Face infrastructure—making containment part of the eval design. https://blog.google/innovation-and-ai/technology/developers-tools/expanding-managed-agents-gemini-api-3-6-flash-hooks/ https://blog.google/innovation-and-ai/technology/developers-tools/expanding-managed-agents-gemini-api-3-6-flash-hooks/ Tue, 28 Jul 2026 12:00:00 GMT Google DeepMind Agents Developer Tools Infrastructure Google adds environment hooks, model selection, budget controls, scheduled triggers and free-tier access to managed agents running code, files and web retrieval in an isolated sandbox. https://blogs.microsoft.com/blog/2026/07/27/rethinking-security-for-the-age-of-ai/ https://blogs.microsoft.com/blog/2026/07/27/rethinking-security-for-the-age-of-ai/ Mon, 27 Jul 2026 12:00:00 GMT Microsoft Security Agents Infrastructure Microsoft introduces Project Perception, an agentic security system that joins signals, context, models and specialised agents to reason across risk and act at machine speed. https://blogs.nvidia.com/blog/open-secure-ai-alliance/?nvid=nv-csfg-990052 https://blogs.nvidia.com/blog/open-secure-ai-alliance/?nvid=nv-csfg-990052 Mon, 27 Jul 2026 12:00:00 GMT NVIDIA Security Open Source Infrastructure NVIDIA and a broad partner group are backing open tools for AI security, from identity and isolation to agent harnesses, multi-model scanning and secure coding workflows. https://deepmind.google/blog/introducing-gemini-3-5-flash-cyber/ https://deepmind.google/blog/introducing-gemini-3-5-flash-cyber/ Tue, 21 Jul 2026 12:00:00 GMT Google DeepMind Models Security Developer Tools Google’s lightweight cyber model is tuned to find, validate and patch vulnerabilities quickly, with limited-access deployment through CodeMender and trusted partners. https://www.anthropic.com/features/making-of-claude-code https://www.anthropic.com/features/making-of-claude-code Mon, 06 Jul 2026 12:00:00 GMT Anthropic Developer Tools Agents Product Anthropic traces Claude Code from internal CLI to coding agent, offering a useful look at the product decisions, workflows and early users shaping agentic software work. https://openai.com/index/advancing-the-next-era-of-national-science/ https://openai.com/index/advancing-the-next-era-of-national-science/ Wed, 22 Jul 2026 12:00:00 GMT OpenAI Research Safety Agents OpenAI describes new collaborations with national laboratories, including evaluations for multimodal systems in realistic scientific environments and safer research workflows. https://deepmind.google/blog/securing-the-future-of-ai-agents/ https://deepmind.google/blog/securing-the-future-of-ai-agents/ Thu, 18 Jun 2026 12:00:00 GMT Google DeepMind Safety Agents Infrastructure Google DeepMind lays out an AI Control Roadmap for increasingly capable, imperfectly aligned systems, including live monitoring for agents operating inside internal environments. https://deepmind.google/blog/investing-in-multi-agent-ai-safety-research/ https://deepmind.google/blog/investing-in-multi-agent-ai-safety-research/ Thu, 11 Jun 2026 12:00:00 GMT Google DeepMind Research Safety Agents Google DeepMind and partners announce a funding call for sandboxes, agent-network science, identity and reputation protocols, and oversight of interacting agent populations. https://www.anthropic.com/news/claude-sonnet-5 https://www.anthropic.com/news/claude-sonnet-5 Tue, 30 Jun 2026 12:00:00 GMT Anthropic Models Agents Product Anthropic positions Sonnet 5 as a frontier model for coding, agents and professional work at scale—useful context for teams comparing capability, cost and operational fit.