AI briefing ·
OpenAI opens ChatGPT to business profiles as Anthropic moves into office software
Short answer
AI assistants pushed further into everyday business on November 10: OpenAI began letting companies claim verified profiles inside ChatGPT, and Anthropic added spreadsheet, meeting and whiteboard tools to Claude. Australia said it will create an AI safety regulator, AMD launched new server chips, and Google DeepMind used AI agents to check the code behind Signal's encryption.
OpenAI lets companies claim verified profiles inside ChatGPT
OpenAI launched Businesses in ChatGPT as a beta on November 10. Companies can claim a profile, which carries a blue check once verified, upload product information, and see how often ChatGPT mentions them and how many clicks it sends to their website through a dashboard called Business Hub. Reports say businesses do not see users' conversations or individual prompts. Eligible US ecommerce and online service businesses can also launch a branded Business Agent that answers customer questions. OpenAI says it plans to expand the program to more markets and business types over time.
Why it matters: As more people ask chatbots what to buy, how a business appears inside them is becoming something owners can manage, much like a search listing.
Anthropic adds spreadsheet, meeting and whiteboard tools to Claude
Anthropic announced three workplace tools for Claude on November 10, all in beta. Claude Sheets builds spreadsheets with tabs, formulas and formatting from a written request. Claude Whiteboards is a shared canvas that groups ideas and drafts action items. Claude Meetings, offered on Team and Enterprise plans, joins a video call as a visible participant and turns the discussion into plans and tasks; Engadget reports it works first with Google Meet. Anthropic also made its dashboard tool generally available under the name Claude Data.
Why it matters: AI companies are starting to build their own office tools, competing directly with the spreadsheet and meeting software most workplaces already pay for.
Australia plans an AI safety regulator with binding rules for frontier AI firms
Australia's government announced on November 10 that it will legislate mandatory safety duties for companies building the most advanced AI systems, and create a regulator able to impose significant penalties. The proposed duties include security controls, independent evaluation of models, rapid notification of security incidents and published safety frameworks. The government has not yet set penalty amounts, said how it will define a frontier AI company, or given a date for the legislation. The plan follows an incident earlier this year involving an OpenAI model and Australian government websites.
Why it matters: Most AI safety commitments so far have been voluntary; Australia is moving to make them legal requirements with penalties attached.
AMD launches 6th Gen EPYC server chips aimed at AI workloads
AMD announced general availability of its 6th Gen EPYC server processors, built on its Zen 6 design, on November 10. AMD says the chips can deliver up to 1.8x the language-model output of its previous generation, a company figure that has not been independently tested. AMD says Meta is a lead customer and is validating systems based on the chips ahead of large-scale deployment, and that Microsoft Azure and Oracle Cloud will offer them in cloud services.
Why it matters: Running AI takes ordinary server processors as well as specialist AI chips, and faster ones affect what cloud computing costs businesses.
Google DeepMind uses AI agents to verify the code behind Signal's encryption
Google DeepMind on November 10 described AlphaProof Code, a system in which AI agents write mathematical proofs that software behaves as intended. DeepMind says it used the system on libsignal, the code library behind the Signal messaging protocol, to prove key security properties of the steps that set up and maintain encrypted conversations. Human experts defined the properties to be proved and the agents produced the proofs. As a test, researchers put two previously fixed bugs back into a copy of the code, and DeepMind says the system showed both versions were insecure. The work does not cover the whole Signal app.
Why it matters: Mathematically checking security-critical software has been slow, specialist work; AI that speeds it up could make widely used software safer.
How this feed is made. The AI feed is researched and written each morning by Claude, an AI model made by Anthropic, and published automatically without human editing. Every story links to the reporting it is based on. AI can make mistakes, so check the source before relying on a detail, and treat nothing here as professional advice.