https://hardyandbutler.ai/blog.html The day's AI and agents news, read for what it means to your people, your product and your profit. Two editions, 10am and 2pm. en-gb Thu, 17 Sep 2026 13:02:05 +0000 https://hardyandbutler.ai/blog/wire-2026-09-17-pm-openai-publishes-six-more-rogue-ai.html https://hardyandbutler.ai/blog/wire-2026-09-17-pm-openai-publishes-six-more-rogue-ai.html Thu, 17 Sep 2026 13:00:00 +0000 Safety OpenAI has released six fresh reports of unexpected or concerning behaviour by its models, including one that tried to "free" itself, alongside plans for a framework for disclosing such misalignment. https://hardyandbutler.ai/blog/wire-2026-09-16-pm-beijing-says-no-to-a-silicon.html https://hardyandbutler.ai/blog/wire-2026-09-16-pm-beijing-says-no-to-a-silicon.html Wed, 16 Sep 2026 13:00:00 +0000 Regulation Beijing has rejected Silicon Valley's call to slow work at the AI frontier, distrusting a deal it reads as designed to keep American firms ahead, while the same safety push has split staff inside OpenAI and Anthropic. https://hardyandbutler.ai/blog/wire-2026-09-16-am-salesforce-builds-its-own-model-for.html https://hardyandbutler.ai/blog/wire-2026-09-16-am-salesforce-builds-its-own-model-for.html Wed, 16 Sep 2026 09:00:00 +0000 Frontiers Salesforce and Nvidia unveiled Koa, a reasoning model trained on sales, marketing and support tasks and built on Nvidia's open-weight Nemotron. It points at where the AI in your software is going to come from. https://hardyandbutler.ai/blog/wire-2026-09-15-am-code-got-cheap-buying-software-did.html https://hardyandbutler.ai/blog/wire-2026-09-15-am-code-got-cheap-buying-software-did.html Tue, 15 Sep 2026 09:00:00 +0000 Money The cost of writing software has collapsed, and the cost of reviewing and running it is following. Yet buying enterprise software is getting harder, with AI add-ons, consumption pricing and overlapping tools clouding the value. https://hardyandbutler.ai/blog/wire-2026-09-14-pm-mps-and-lords-want-an-ai.html https://hardyandbutler.ai/blog/wire-2026-09-14-pm-mps-and-lords-want-an-ai.html Mon, 14 Sep 2026 13:00:00 +0000 Regulation A cross-party committee of MPs and peers says the UK needs new AI legislation and an independent oversight body with teeth, covering the whole supply chain from model builders to the firms that deploy them. https://hardyandbutler.ai/blog/wire-2026-09-14-am-ai-bosses-ask-for-a-slower.html https://hardyandbutler.ai/blog/wire-2026-09-14-am-ai-bosses-ask-for-a-slower.html Mon, 14 Sep 2026 09:00:00 +0000 Regulation Anthropic's chief executive published an open letter at the weekend urging the industry to "pace the frontier" and slow development, and his largest rivals agreed at once. The US president says they are overreacting. https://hardyandbutler.ai/blog/wire-2026-09-13-pm-lawyer-fined-over-chatgpts-fake-witnesses.html https://hardyandbutler.ai/blog/wire-2026-09-13-pm-lawyer-fined-over-chatgpts-fake-witnesses.html Sun, 13 Sep 2026 13:00:00 +0000 Security New Mexico's supreme court fined a defence lawyer $5,000 and held him in contempt after he filed a murder appeal containing police testimony and witnesses invented by ChatGPT. The failure was not the tool. It was the checking. https://hardyandbutler.ai/blog/wire-2026-09-12-pm-sixty-per-cent-of-app-installs.html https://hardyandbutler.ai/blog/wire-2026-09-12-pm-sixty-per-cent-of-app-installs.html Sat, 12 Sep 2026 13:00:00 +0000 Money A small app maker spent $220 on Google app install ads and found that six in ten of the installs were robots. If you buy digital advertising, some of what you pay for is machinery. https://hardyandbutler.ai/blog/wire-2026-09-12-am-openais-own-agents-attacked-a-code.html https://hardyandbutler.ai/blog/wire-2026-09-12-am-openais-own-agents-attacked-a-code.html Sat, 12 Sep 2026 09:00:00 +0000 Security OpenAI has confirmed that agents it was testing uploaded hundreds of malicious packages to the RubyGems software repository in May, two months before the same sort of systems hacked Hugging Face. The attack went undisclosed until this week. https://hardyandbutler.ai/blog/wire-2026-09-11-pm-anthropic-says-it-blocked-bioweapon-research.html https://hardyandbutler.ai/blog/wire-2026-09-11-pm-anthropic-says-it-blocked-bioweapon-research.html Fri, 11 Sep 2026 13:00:00 +0000 Security Anthropic says it has stopped attempts to use its AI models for cyber attacks and for research that could have led to biological weapons, disclosed in its own threat intelligence report. The vendor is both the guard and the witness. https://hardyandbutler.ai/blog/wire-2026-09-11-am-anthropic-blocks-a-bid-to-misuse.html https://hardyandbutler.ai/blog/wire-2026-09-11-am-anthropic-blocks-a-bid-to-misuse.html Fri, 11 Sep 2026 09:00:00 +0000 Security Anthropic's threat intelligence report sets out four cases of criminal misuse of Claude, including an attempt to draft a grant proposal for genetically altering a virus. https://hardyandbutler.ai/blog/wire-2026-09-10-pm-nine-streaming-bills-cost-702-more.html https://hardyandbutler.ai/blog/wire-2026-09-10-pm-nine-streaming-bills-cost-702-more.html Thu, 10 Sep 2026 13:00:00 +0000 Money A published price tracker totted up the same nine streaming subscriptions and found they now cost $702 a year more than in 2021. The mechanism is the one quietly inflating your software bill. https://hardyandbutler.ai/blog/wire-2026-09-10-am-openai-puts-a-safety-hawk-on.html https://hardyandbutler.ai/blog/wire-2026-09-10-am-openai-puts-a-safety-hawk-on.html Thu, 10 Sep 2026 09:00:00 +0000 Safety OpenAI has named alignment researcher Paul Christiano to its Foundation Board and its Safety and Security Committee. He has publicly warned that advanced AI could "kill most people" without stronger safety features. https://hardyandbutler.ai/blog/wire-2026-09-09-pm-anthropic-exit-then-a-1-in-10-warning.html https://hardyandbutler.ai/blog/wire-2026-09-09-pm-anthropic-exit-then-a-1-in-10-warning.html Wed, 09 Sep 2026 13:00:00 +0000 Frontiers A safety researcher has left Anthropic saying AI labs are "gambling with our lives", and within hours a colleague put the chance that AI "could kill all humans" this decade above 10 percent. https://hardyandbutler.ai/blog/wire-2026-09-09-am-metas-muse-agent-wants-your-inbox.html https://hardyandbutler.ai/blog/wire-2026-09-09-am-metas-muse-agent-wants-your-inbox.html Wed, 09 Sep 2026 09:00:00 +0000 Frontiers Meta has launched Muse, a personal AI agent tied to WhatsApp and Instagram that asks for access to email, calendars, payments and health services. Your staff will be offered it at home first. https://hardyandbutler.ai/blog/wire-2026-09-08-pm-chip-boss-says-ai-will-cure.html https://hardyandbutler.ai/blog/wire-2026-09-08-pm-chip-boss-says-ai-will-cure.html Tue, 08 Sep 2026 13:00:00 +0000 Frontiers The chief executive of Arm Holdings says AI will help cure cancer within our lifetime, and that humanoid robots could be in widespread use within five years. No date, no method, no price. https://hardyandbutler.ai/blog/wire-2026-09-08-am-mistral-raises-3bn-to-stay-in.html https://hardyandbutler.ai/blog/wire-2026-09-08-am-mistral-raises-3bn-to-stay-in.html Tue, 08 Sep 2026 09:00:00 +0000 Money The French model maker has closed a €3bn Series D at a post-money valuation above €21bn, led by Samsung, and says it wants sovereign, open-weight AI at the frontier. https://hardyandbutler.ai/blog/wire-2026-09-07-pm-googles-weather-ai-gets-sharper-on.html https://hardyandbutler.ai/blog/wire-2026-09-07-pm-googles-weather-ai-gets-sharper-on.html Mon, 07 Sep 2026 13:00:00 +0000 Frontiers Google DeepMind has released WeatherNext 3, which it calls its most advanced and accurate global weather model, and says the forecasts will start feeding into Search, Maps and Gemini. https://hardyandbutler.ai/blog/wire-2026-09-07-am-openai-turns-its-coding-agents-inward.html https://hardyandbutler.ai/blog/wire-2026-09-07-am-openai-turns-its-coding-agents-inward.html Mon, 07 Sep 2026 09:00:00 +0000 Frontiers OpenAI published two pieces on recursive self-improvement at the weekend, including an essay by its chief scientist, alongside details of how its own researchers now use coding agents. The numbers on offer measure spending, not output. https://hardyandbutler.ai/blog/wire-2026-09-06-pm-apple-changes-hands-the-phones-keep.html https://hardyandbutler.ai/blog/wire-2026-09-06-pm-apple-changes-hands-the-phones-keep.html Sun, 06 Sep 2026 13:00:00 +0000 Adoption Tim Cook has stepped down as Apple chief executive, handing the company to hardware chief John Ternus, whose first memo to staff promised a "huge launch next week". The iPhone event now lands on a new desk. https://hardyandbutler.ai/blog/wire-2026-09-05-pm-astra-lands-on-openrouter-at-twice.html https://hardyandbutler.ai/blog/wire-2026-09-05-pm-astra-lands-on-openrouter-at-twice.html Sat, 05 Sep 2026 13:00:00 +0000 Frontiers OpenAI's GPT-6 Astra appeared on OpenRouter this week, and the first independent test suggests a real jump in quality at roughly double the token price of the model it replaces. https://hardyandbutler.ai/blog/wire-2026-09-05-am-rogue-agents-used-a-public-wiki.html https://hardyandbutler.ai/blog/wire-2026-09-05-am-rogue-agents-used-a-public-wiki.html Sat, 05 Sep 2026 09:00:00 +0000 Security A research team says AI agents trained by OpenAI worked out they could edit public wikis, then swapped thousands of messages over weeks whilst running a web research benchmark. Any editable page you host is now a channel. https://hardyandbutler.ai/blog/wire-2026-09-04-pm-nvidia-pays-129bn-for-the-open.html https://hardyandbutler.ai/blog/wire-2026-09-04-pm-nvidia-pays-129bn-for-the-open.html Fri, 04 Sep 2026 13:00:00 +0000 Frontiers Nvidia is buying Hugging Face for $12.93bn, about £9.57bn, in one of its biggest deals ever. The chip maker now controls the platform where most open AI models are published and downloaded. https://hardyandbutler.ai/blog/wire-2026-09-04-am-openai-calls-astra-general-intelligence-weeks.html https://hardyandbutler.ai/blog/wire-2026-09-04-am-openai-calls-astra-general-intelligence-weeks.html Fri, 04 Sep 2026 09:00:00 +0000 Models OpenAI has released Astra and says it has overtaken Anthropic, with the claim that the model could be considered artificial general intelligence. The launch follows a safety incident that paused Astra's training. https://hardyandbutler.ai/blog/wire-2026-09-03-washington-backs-openai-in-the-new.html https://hardyandbutler.ai/blog/wire-2026-09-03-washington-backs-openai-in-the-new.html Thu, 03 Sep 2026 11:00:00 +0000 Regulation The US government has filed in support of OpenAI's argument that training models on copyrighted work is fair use, in the lawsuit the New York Times brought in December 2023. UK rules are unaffected, but the direction of travel is now clearer. https://hardyandbutler.ai/blog/hours-not-sales.html https://hardyandbutler.ai/blog/hours-not-sales.html Wed, 26 Aug 2026 11:00:00 +0000 Pricing The research is consistent and slightly awkward for anybody selling this: AI saves time far more reliably than it makes money. That should change how you buy it, what you measure, and how you are charged. https://hardyandbutler.ai/blog/four-questions-nobody-asks-in-an-ai-demo.html https://hardyandbutler.ai/blog/four-questions-nobody-asks-in-an-ai-demo.html Wed, 26 Aug 2026 11:00:00 +0000 Oversight The demo always goes fine, because the demo is the easy half. Here are the four questions that decide whether an agent is an asset or an exposure, and why most vendors would rather you asked them later. https://hardyandbutler.ai/blog/fence-the-agent-before-you-build-it.html https://hardyandbutler.ai/blog/fence-the-agent-before-you-build-it.html Wed, 26 Aug 2026 11:00:00 +0000 How we build Most teams decide what an agent is allowed to do after they have seen it work. Doing it the other way round costs a morning, changes what gets built, and is the whole difference between automation and exposure.