The top things worth knowing about in AI today.
Harvard researchers Fiona Chen and James Stratton analysed about 300 million work records from 718 companies between 2021 and March 2026. Semi-autonomous coding agents lifted lines of code by 30%, commits by 20% and pull requests by 23%, but average review time rose 49% and overall output did not change significantly. If your team has adopted coding agents, review capacity is now the constraint to plan around.
Read moreLumen's Black Lotus Labs has tracked a campaign called PoeLLM since April that has compromised more than 3,400 servers, most of them running self-hosted AI tools such as LiteLLM and Ollama. It exploits an exposed LiteLLM endpoint (CVE-2026-42271) to install crypto miners and turn victims into scanners for further attacks. If you self-host a model gateway, patch it, keep ports 3000 and 4000 off the public internet, and put AI tooling in your normal patch cycle.
Read moreHCLTech has signed a Global Premier Partnership with Anthropic, setting up a dedicated Anthropic business unit and a centre of excellence for agentic AI. Its services teams will build industry solutions on Claude, Cowork and Claude Code, and Claude will be built into HCLSoftware products including BigFix and AppScan. With more than 223,000 staff in 60 countries, HCLTech joins a growing list of large IT services firms standardising client AI work on a single model provider.
Read moreButterfly Effect, the company behind the Manus AI agent, raised more than $500 million in a round led by Boyu Capital and IDG Capital, with Tencent and HSG also taking part. It is the first raise since Chinese authorities ordered Meta's $2 billion acquisition unwound in April. Its new Cue app gives personal agents their own email address, phone number and wallet with user-set spending limits, a useful signal of where consumer agent products are heading.
Read moreUK AI minister Kanishka Narayan says Britain cannot rely entirely on tech companies to keep AI safe, as the government moves to strengthen national cyber defences. The speech follows a security incident at OpenAI that raised concerns adversaries could steal advanced models, and comes as other governments push for tighter rules. Organisations buying AI services should expect more pressure on suppliers to demonstrate their security controls, not just describe them.
Read moreDeepSeek has led a funding round in a Hangzhou robot AI startup founded by a former senior Tencent scientist, valuing it at about $400 million, according to Bloomberg reporting. It follows DeepSeek's earlier move into robotics through Unitree. China's leading model labs are now putting capital into physical AI, which is worth watching if you track where low-cost open models will show up next.
Read moreA short read every week — the few things worth your time, and nothing that isn’t.