This morning’s briefing opens with the FCA’s plan to use AI to automate financial regulation. Among the tools, GitHub Copilot adds deeper review for complex work, while Grok’s voice API has automatically upgraded agents using its “latest” model setting.
New releases
Codex. GPT-5.5 retired from Codex for ChatGPT sign-ins at 20:00 BST on 14 October; replacements are GPT-6.1 Sol on eligible paid plans or GPT-6 Luna for Free and Go desktop users, where available, while API-key sessions are unaffected. More
Claude Code desktop. A 14 October hands-on report shows Claude Desktop v2.37337.1 adding saved, named environments for local, cloud and SSH Code sessions; Code access requires Claude Pro, Max, Team or Enterprise. More
Claude Platform. A new beta lets teams using Claude Platform export request and tool-call traces, including token and timing data, to OpenTelemetry monitoring tools; API tracing requires an administrator to configure a destination in Claude Console. More
GitHub Copilot. HydraFusion’s research preview adds Max, deeper review and a model pool including Claude Opus 5.5, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, for Copilot Pro, Pro+, Business and Enterprise users with previews enabled in VS Code, the Copilot app or CLI. More
Visual Studio Code. VS Code 1.142 now routes full chats governed by configured permission or sandbox policies through Copilot Agent Host, excluding Claude and Codex sessions while those policies apply unless an administrator opts out; Windows terminal sandboxing also moves from Local sessions to Agent Host. More
Visual Studio Code and GitHub Copilot. VS Code 1.142 adds preview canvases, including shared Kanban boards, for local desktop sessions using the Copilot option; Copilot also begins becoming the default for new sessions and restores a read-only Ask mode. More
Grok Voice Agent API. Grok Voice 3.0 launched on 14 October at unchanged API rates of US$0.08 per minute plus US$0.004 per text input, and agents using grok-voice-latest have already switched automatically. More
Elsewhere in AI
The UK’s Financial Conduct Authority plans to use AI to automate financial regulation as it prepares its response to frontier models, the Financial Times reports. Read
Eight US states have sued OpenAI over alleged harm to children, seeking changes to its products for young users, The Guardian reports. Read
Anthropic told investors that quarterly revenue rose 50%, although growth was slower than some expected amid market anxiety about AI, the Financial Times reports. Read
OpenAI alleges that an employee sent confidential information about its unreleased AI device to Apple, as the companies face off in court, the Financial Times reports. Read
Nvidia pledged US$250m in computing power over five years for UK AI safety research amid growing concerns about autonomous agents, The Guardian reports. Read