Skip to content
IIAIT

AI Today · 3 November 2026

ChatGPT and Codex move to one shared allowance, and an infostealer that asks AI to rank its victims

OpenAI tells users that chat, agent work and Codex will draw on a single usage allowance, Cisco Talos analyses Windows malware that hands stolen data to commercial AI models for appraisal, and Anthropic's updated listing filing warns that AI may soon be speeding up its own research.

  1. 01

    OpenAI will merge ChatGPT's Chat and Work tabs and give ChatGPT and Codex one shared usage allowance

    In an email to users reported on 2 November, OpenAI said it will combine the Chat and Work tabs of the ChatGPT desktop app into a single experience, tested first with a small group of users before a wider rollout; Codex stays a separate part of the app. ChatGPT and Codex will also draw on one shared usage allowance instead of separate limits. OpenAI says most users will not wait for resets more often than today, but advanced models and complex tasks use more of the allowance, while Instant chats do not count towards it. Separately, an in-app notice says the older ChatGPT Classic desktop app will be retired on 30 January 2027.

    What it means for you

    If you or your team use both ChatGPT and Codex on one plan, heavy coding or agent sessions may now reduce what is left for everyday chat. Keep an eye on your usage in the app settings once the change arrives, and plan the move off ChatGPT Classic before the end of January.

  2. 02

    Cisco Talos: VOLTA infostealer sends stolen data to Claude, Kimi or Gemini to rank victims by value

    Cisco Talos has analysed VOLTA, a Windows infostealer that collects browser passwords and cookies, payment card data, Microsoft OAuth tokens, SSH and cloud keys, Slack and Outlook data and crypto wallets. It then sends a profile of each victim through OpenRouter to Anthropic's Claude Opus 4.6, asking for a ranked assessment of what is worth exploiting; if the model refuses, it falls back to Moonshot's Kimi K2 and then Google's Gemini 2.5 Flash. Talos says the AI step replaces the manual sorting of stolen data rather than adding new capabilities, and it did not report seeing the AI output used in real incidents. The malware deletes itself and appears to be sold as a builder kit.

    What it means for you

    The main risk is still the theft itself, so the usual defences matter most: keep credentials in hardware-backed stores, use phishing-resistant sign-in, and revoke tokens and change passwords quickly after any suspected infection. Talos also suggests blocking or alerting on Chrome being launched in headless mode by unknown programs.

  3. 03

    Anthropic's amended listing filing says AI may soon dramatically accelerate its own research

    In an amended S-1 registration statement filed with the US Securities and Exchange Commission on 2 November, Anthropic added a risk factor saying its internal estimates for future models suggest it may soon meet its Responsible Scaling Policy's threshold of dramatically accelerated progress attributable to AI. The company says no current model has been designated as crossing that threshold, that in the past it has recognised such acceleration only after it happened, and that it does not expect to safely let Claude fully autonomously design and build new frontier models. It says it is considering updates to the policy to keep humans in control. Commentators note that a listing filing is also a document aimed at investors.

    What it means for you

    Nothing changes in the Claude products you use today. It is a signal worth following: the pace of model releases may increase, so plan regular re-testing of the AI tools you depend on and keep human review in place for important outputs.