From the archive
AI agents use language models and software tools to plan tasks, write code and carry out multi-step work. Follow the latest AI agent news, coding assistants and automation frameworks, with attention to reliability, permissions and human oversight. Read the source-linked reports to understand what each system can actually do, where it still fails and what changes for developers and businesses.
Lead story
Google’s AI infrastructure chief Amin Vahdat says power is the most fundamental constraint on scaling AI, even as failures, networking and hardware efficiency all put pressure on large systems. His account also shows why faster chips alone will not settle the question: the useful measure is how much…
Read the report & discussion ↗Watch Desk
Wajo is building an AI agent that can contact people and businesses, arrange services and pay on a user’s behalf. The startup’s pitch is that trust should be built into the product from the start, which matters when an assistant is being given access to your inbox, contacts…
Meta AI Watch
Meta has opened Muse Gadgets, a project for developers to connect its personal AI agent to home-built hardware. The useful shift is from an assistant that lives in an app to one that could reach buttons, screens and sensors around the home.
METR Watch
METR says a vulnerability in the Inspect transcript viewer could have let an AI agent change what a human reviewer saw, including earlier actions in the transcript. The flaw was patched within a day of being reported, and METR says it has not observed agents exploiting it.
Watch Desk
Utah Governor Spencer Cox has signed an order setting out how state agencies should bring AI into government. It assigns responsibilities for staff training, cybersecurity safeguards and human review of automated decisions affecting individual rights, KSL reports.
Watch Desk
AI agents can help sports fans track tickets, manage calendars and weigh up fantasy trades, according to a Sportico writer’s hands-on tests. The experiments mostly worked, but they also exposed the awkward trade: convenience may mean handing an agent access to personal accounts, and trusting answers that can…
Microsoft AI Watch
Microsoft has opened its Frontier Partner specialisation to partners delivering AI agents, offering qualified firms co-selling and other commercial support. The designation covers work across Microsoft’s cloud, productivity, development and security products, making it a notable push to turn agent expertise into customer services.
Watch Desk
Meta and Sierra have announced an open standard for personal AI agents interacting with businesses, with authentication and consumer control at its centre. The plan brings a notable group of commerce and technology companies together, but the announcement does not yet set out a technical specification or delivery…
Watch Desk
AI customer-service startup Siena has raised $17 million to expand its “Agent of Record” platform, which brings purchase history and customer conversations together for support and shopping agents. The pitch is continuity across channels; the practical question is how well that shared context works when the information is…
Watch Desk
Supabase has announced $150 million in new funding and the acquisition of database company Turso, betting that a surge in software built with AI agents will need a great many more databases. Supabase says it is adding more than one million users and four million databases a month,…
Watch Desk
Google Cloud and Mysten Labs have announced a system designed to help businesses check whether AI agents stayed within their authorised powers. The proposal combines private activity logs with cryptographic proofs, addressing a practical question as agents are asked to take actions on companies’ behalf: how can anyone…
Watch Desk
IBM and Red Hat say their Lightwell programme has fixed more than 400 previously unknown vulnerabilities in widely used Java libraries and opened a service for enterprise customers to request priority reviews and backported fixes. The announcement also ties the effort to risks from autonomous AI agents, though…
Watch Desk
Enterprise AI is shifting from deploying individual tools to organising teams of agents, argues NTT Research chief executive Kazuhiro Gomi. His case is that leaders will need to decide how agents divide work, learn their roles and remain under human oversight, not simply pick a model and let…
Databricks Watch
Databricks has made Meta’s ads MCP server available in its Marketplace, connecting Meta campaign tools with business data and models in Databricks. Marketers can use Genie to ask questions about campaign performance and take actions such as adjusting budgets or creating campaigns, with administrative controls around what the…
Watch Desk
AI agents could one day guide people through legal problems, draft claims and negotiate with institutions, the Law Society warns. Its foresight study argues that accountability and meaningful human oversight need settling before such systems become a doorway into justice.
AMD Watch
AI agents may be shifting some of the computing workload back towards CPUs, giving AMD and Intel a larger role alongside the GPUs that have dominated the AI boom. The change matters because agents need processors to keep their virtual computers and tasks running, not just chips to…
OpenAI Watch
OpenAI and Synopsys are developing a specialised AI model that can operate chip-design software, with agents set to run and refine engineering workflows for human review. It is a substantial step towards putting AI to work inside semiconductor design, though there is no launch date or pricing yet.
Watch Desk
Deutsche Telekom says AI and automation could cut its indirect costs by about €2.5 billion by 2030, while also growing revenue from AI-related services. The company is already using AI in customer service and network operations, giving its ambitious savings target a few concrete numbers to answer to.
Shopify Watch
Shopify says 65% of shoppers plan to use AI for at least one shopping task this holiday season, but only one-third say they trusted an agent to buy on their behalf. For merchants, the immediate opportunity is helping AI recommend the right product, not assuming shoppers are ready…
Cohere Watch
Cohere and Thales say a new task under a Canadian naval support contract will bring agentic AI into maintenance and support activities for Royal Canadian Navy vessels. The planned system is aimed at routine work, with personnel remaining in control and able to focus on more complex tasks.
Watch Desk
A security report describes vulnerabilities that let attackers turn trust between AI agents into a route for malicious instructions, including a Google MCP toolbox flaw that could direct requests to internal network endpoints. The practical lesson for organisations using connected agents is blunt: a tool call passed through…
