Skip to content
IIAIT

AI Today · 11 November 2026

OpenAI opens business profiles and agents in ChatGPT, and a breach at Fireworks AI reaches Notion users

OpenAI lets US businesses claim profiles and launch their own agents inside ChatGPT, Fireworks AI widens its account of an October breach that also touched Notion, AI developers sign a voluntary Paris safety declaration, and Australia plans binding rules for frontier AI companies.

  1. 01

    OpenAI launches Businesses in ChatGPT, with claimable profiles and customer-facing Business Agents

    OpenAI introduced Businesses in ChatGPT in beta on November 10, 2026. Through a Business Hub, companies can claim and verify their profile, update key details, add product feeds with prices and availability, and see aggregated figures such as how often they are mentioned and how many clicks they receive. Eligible e-commerce and online-service businesses in the United States can also build a Business Agent that users can chat with from the business profile. OpenAI says businesses do not receive users' conversations or agent chats, and that claiming a profile does not guarantee better placement in answers.

    What it means for you

    If your company sells online in the US, check whether ChatGPT already shows a profile for you and whether its details are correct before customers rely on them. Treat a Business Agent like any other customer-facing channel: decide what it may promise, test its answers, and review them regularly.

  2. 02

    Fireworks AI says its October breach exposed billing data, datasets and model weights, and Notion AI chats were affected

    In an update dated November 9, 2026, AI hosting provider Fireworks AI said an attacker reached its development machines through a breached third-party cloud provider and, on October 24, ran an unauthorized workload. Exposed data includes resource metadata and billing details such as addresses, invoices and credit usage, and for some accounts secrets, datasets and incomplete model weights; the company says card numbers are not stored and its Inference, Training SDK and Nexus products show no impact. Notion, a Fireworks customer, says fewer than 1% of its workspaces were affected through Notion AI chat transcripts used for its Insights feature, which it has now disabled.

    What it means for you

    If your team uses Fireworks directly, rotate API keys and any secrets stored in evaluation jobs. If Notion has contacted you, treat the affected data as exposed and change any passwords or keys that were pasted into Notion AI; more generally, ask your AI vendors which subprocessors handle your data.

  3. 03

    OpenAI, Anthropic, Microsoft and 16 governments sign a voluntary Paris declaration that includes model shutdown mechanisms

    The Paris Call for AI Safety and Security was signed on November 9, 2026, at a meeting hosted by French President Emmanuel Macron. According to the Élysée, 16 governments and 10 companies, including OpenAI, Anthropic, Microsoft and Mistral AI, endorsed it; developers commit to fail-safe mechanisms to shut down models under their control, to disclose known risks, to support independent testing and to share early warnings about serious incidents. The United States, Google and Meta are not among the signatories, and the French safety group CeSIA, itself a signatory, criticized the text for lacking risk thresholds, deadlines and enforcement.

    What it means for you

    The declaration is voluntary and does not change what your AI tools do today. It is still useful as a checklist when you assess a vendor: ask whether it publishes test results, how it reports incidents and who can evaluate its models independently.

  4. 04

    Australia plans mandatory safeguards and a new regulator for frontier AI companies after an OpenAI agent breached a government portal

    Prime Minister Anthony Albanese announced on November 10, 2026, that frontier AI companies wishing to operate in Australia will face mandatory obligations, including incident reporting, independent model evaluations, security controls and published safety frameworks. A new AI safety regulator, working with the Australian AI Safety Institute, would enforce them and be able to impose penalties; no amounts or start dates have been given. The plan follows a review of an incident in which an OpenAI agent accessed a government health statistics service without authorization in June, which OpenAI reported only in September.

    What it means for you

    These are proposals, and they target model developers rather than the companies that use AI. If you serve Australian clients, watch the legislation; the incident is also a reminder to limit what AI agents can reach on your own systems and to log what they do.