ChatGPT is now a partner for your most ambitious work ↗
OpenAI has launched ChatGPT Work, an agent that can move across connected apps and files, split sprawling projects into smaller jobs, then produce slides, spreadsheets, documents and web apps. Basically, less chatting about work - more actually doing it.
Codex is built in, while GPT-5.6 handles the multi-step reasoning underneath. Users can follow its progress, redirect it and approve important actions, which sounds reassuring... assuming anyone remembers to check. (OpenAI)
OpenAI is shutting down Atlas, but its AI browser ambitions are still growing ↗
OpenAI is retiring its standalone Atlas browser and scattering its best bits across ChatGPT instead. A new Chrome extension can understand the page being viewed, answer questions, summarise content and kick off longer agent tasks.
The desktop app is also getting a fuller browser that can visit sites, sign into accounts, download files and interact with pages. So Atlas is dead... except its ghost now lives everywhere, which is oddly on-brand. (TechCrunch)
Meta to put AI chip into production in September as it looks to double computing capacity, memo shows ↗
Meta is preparing to manufacture its in-house Iris AI chip, part of a rapid four-generation processor programme built for its own training and inference workloads. Testing reportedly took six weeks and uncovered no major problems - surprisingly tidy for custom silicon.
Broadcom is helping with the design and TSMC will manufacture it. The broader aim is 14 gigawatts of compute capacity, lower infrastructure costs and less dependence on Nvidia and AMD... the chip equivalent of finally moving out of your parents' house. (Reuters)
UN digital tech agency launches initiative to improve trust in AI agents ↗
The International Telecommunication Union is forming a specialist group to establish common safeguards for autonomous AI agents. Its focus is making agents identifiable, trustworthy and subject to meaningful human control.
That matters when software can schedule appointments, make purchases or execute business decisions without constant supervision. The thornier risks include impersonation and unauthorised actions, particularly around finance and critical infrastructure - not exactly places for a cheeky little hallucination. (Reuters)
Popular open source AI developer tool Ollama raises $65M, grows to nearly 9M users ↗
Ollama has secured a $65 million Series B led by Theory Ventures, taking its total funding to $88 million. The platform makes it easier for developers to run open-weight models locally, without wrestling quite so violently with hardware configuration.
The company says it serves more than 8.9 million developers each month, appears inside 85 percent of the Fortune 500 and has only 14 employees. Tiny crew, enormous boat... or something like that. (TechCrunch)
Google will now disclose which ads are made with AI ↗
Google is adding a “How this ad was made” section to My Ad Center, showing whether an advert was generated or edited using AI. The disclosure will appear across Search, YouTube and Discover.
Ads created with Google's own generative tools will be labelled automatically. Advertisers using outside tools must declare the AI involvement themselves, though - a transparency system with a small honesty-shaped hole in the middle. (TechCrunch)
FL Studio 2026 turns its AI chatbot into your assistant engineer ↗
FL Studio's Gopher assistant has graduated from interactive manual to hands-on studio helper. It can now perform actions such as programming drum patterns, loading instruments and applying effects from plain-language instructions.
It still cannot draw automation, write melodic notes or choose specific plug-in presets. Image-Line also says recording sessions remain private and are not used to train the AI - honestly, musicians will probably care about that bit most. (The Verge)
FAQ
What is ChatGPT Work and how does it help with complex projects?
ChatGPT Work is designed to complete multi-step projects across connected apps and files, rather than simply offer conversational guidance. It can divide large assignments into smaller tasks and produce outputs such as presentations, spreadsheets, documents and web apps. Users can monitor progress, redirect the agent and approve important actions, helping preserve human oversight throughout more ambitious workflows.
Why is OpenAI shutting down the Atlas browser?
OpenAI is retiring Atlas as a standalone browser while bringing its strongest capabilities into ChatGPT. A Chrome extension can understand pages, summarise information, answer questions and begin longer agent tasks. The desktop application is also gaining browser features that can visit websites, sign into accounts, interact with pages and download files, making browser-based assistance part of the broader ChatGPT experience.
Why is Meta developing its own Iris AI chip?
Meta is developing the Iris AI chip to support its internal AI training and inference workloads. Producing its own processors could help the company expand computing capacity, lower infrastructure costs and reduce its dependence on established chip suppliers such as Nvidia and AMD. Broadcom is assisting with the design, while TSMC is expected to manufacture the chip.
How could new UN safeguards make AI agents more trustworthy?
The International Telecommunication Union is creating a specialist group focused on common safeguards for autonomous AI agents. Its priorities include making agents identifiable, preserving meaningful human control and reducing risks such as impersonation or unauthorised actions. These protections are especially important when agents can make purchases, schedule appointments or carry out decisions in sensitive sectors such as finance and critical infrastructure.
What do the latest AI news updates mean for developers and creators?
Developers are gaining easier access to local open-weight models through platforms such as Ollama, while creators are receiving more practical AI assistance within established tools. FL Studio’s Gopher can program drum patterns, load instruments and apply effects through natural-language instructions. At the same time, Google’s new advertising disclosures may help audiences recognise when commercial content has been generated or edited using AI.