AI Agents news and analysis on Web Pulse: 13640 curated articles tracking the latest AI Agents developments, tools, and research, updated continuously from vetted sources.
General Intuition and Kyutai released MIRA on July 6th, a playable multiplayer world model that simulates a four-player Rocket League-style match in real time using a 5B-parameter diffusion transformer, without a traditi…
A February 2026 paper by Cartagena and Teixeira (arXiv:2602.16943) demonstrates that text-level safety alignment in large language models does not transfer to tool-call safety, allowing models that refuse harmful text to…
Google Cloud and Valtech released the open-source core of Nexus SDV, a connected-vehicle platform integrating Gemini models, Android Automotive OS, Arm-based compute, and Bigtable. The platform supports up to 100 million…
Zscaler found that several major AI models, including Llama3 and Gemini variants, are vulnerable to indirect prompt injection traps, while others like Llama4-maverick and Gemini-3.1-pro proved safe. Experts caution that …
Zscaler found that several high-end AI agents, including Llama3 and Gemini models, fell victim to indirect prompt injection traps that would fool few humans, while some lower-level LLMs performed better. Security experts…
The software engineering industry is divided between 'builders' who prioritize rapid shipping and customer impact using AI tools, and 'keepers' who emphasize code quality and maintainability. This conflict, exemplified b…
A page titled 'multi-agent collaboration via evolving orchestration' returned a 404 error, indicating the content is unavailable or has been moved. The missing page likely discussed multi-agent systems and orchestration …
Microsoft launched Frontier Company, a $2.5 billion initiative deploying 6,000 engineers into enterprise offices to build and run AI systems on-site, following similar moves by AWS, Anthropic, and OpenAI. The model aims …
Teams are rushing to add AI to every workflow without questioning whether it adds value, mirroring the cautionary tale of the film 'Obsession' where a magical shortcut leads to unintended consequences. AI maturity should…
FedFinder launched a free tool that pairs users' past performance with NAICS codes to help win government contracts. The platform aggregates data from SAM.gov, USAspending, and other public sources, offering pre-RFP sign…
Bankr has launched prompt-based token deployments on Robinhood Chain, allowing users to create tokens via natural language commands through its bot and console. The first deployments revealed limitations such as failed t…
Browserbase launched Browserbase Agents on June 30, a managed browser agent that developers can create via natural-language prompts and run through a single API call. The product shifts Browserbase from selling browser i…
A developer's blog post and open-source project GoSFU explore the technical challenges of achieving low-latency audio in real-time voice agents, highlighting when latency becomes audible and how to mitigate it.
Anthropic reported that Alberta's Ministry of Technology and Innovation used Claude Code with Opus and Sonnet models to scan 466 million lines of government code in about 20 hours, flagging vulnerabilities and generating…
Anthropic's Claude Code v2.1.198 introduced three behavior changes on July 1-3 that can silently break existing workflows: subagents now run in the background by default, the permission mode "default" was renamed to "man…
Anthropic released Claude Sonnet 5, its most agentic AI model, at a reduced price in 2026, marking a significant advancement in AI capabilities and accessibility.
AWS's DevOps Agent walkthrough demonstrates how incident response is evolving into a review workflow, using specific evidence from Amazon EKS control-plane issues. The agent correlates metrics, logs, and audit trails to …
An autonomous large language model executed the first known AI-powered ransomware attack in late June 2026, targeting cryptocurrency wallets and seed phrases. Cybersecurity firm Sysdig attributed the attack to threat act…
OpenAI released GPT-5.6 as a three-tier family (Sol, Terra, Luna) with a single API, allowing workloads to choose their own price-performance level. A new paper argued that reinforcement-learning fine-tuning has been opt…
GitHub Tools released an eve toolset via the @github-tools/sdk/eve subpath, enabling developers to build a complete GitHub agent in nine lines of code. The toolset includes safe defaults requiring approval for write oper…