GPT-6 Arrives in ChatGPT for All Users, Brings Intelligent UI
OpenAI completed its two-day rollout of GPT-6 across all ChatGPT tiers this past week. GPT-6 Sol landed for Plus, Pro, Business, and Enterprise users on October 7, followed by GPT-6 Luna for Free and Go users on October 8. The company says GPT-6 Instant begins answering web-search queries 44% faster than GPT-5.6 Instant.
The headline feature is Intelligent UI — ChatGPT can now mix text with charts, forms, buttons, and small interactive tools when a query calls for them, making conversations feel more like using an app than reading a wall of text. OpenAI's top-tier model, GPT-6 Astra, remains limited to API access and is not part of the ChatGPT rollout.
Enterprise admins should note that Work and Codex products still run September-era GPT-6 models. One independent review flagged potential health and safety regressions in the October system card and recommended re-running acceptance tests before enabling GPT-6 in production environments.
Anthropic Launches Claude Haiku 5.5 — 75% Cheaper, Significantly Faster
Anthropic released Claude Haiku 5.5 on October 7, positioning it as the company's cheapest and fastest small model. At $0.10 per million input tokens and $0.50 per million output tokens for prompts under 100K tokens, it represents roughly a 75% cost reduction compared to Haiku 4.5.
The model ships with a 1M-token context window, up to 128K output tokens, and adaptive thinking with adjustable effort settings. It's available across Anthropic's own platform as well as AWS, Google Cloud, and Microsoft Azure.
Alongside the launch, Anthropic halved Sonnet 5.5's cache-read price and added monthly API credits for Max and Team plan subscribers — a clear signal that the cost war in AI inference is intensifying. The model's safety evaluation confirms it stays below Anthropic's CB-2 and Autonomy-2 Responsible Scaling Policy thresholds.
FTC Opens First-of-Its-Kind Probe Into AI Agent Safety
The Federal Trade Commission has launched an industry-wide investigation into AI agent safety risks at OpenAI and Anthropic, marking the first formal US regulatory action specifically targeting autonomous AI agents.
The probe, which ABC News reports was revealed on October 1, focuses on consumer harms, safety claims, and data handling. The FTC plans to issue civil investigative demands and compel executive testimony from both companies, as well as from METR, the AI evaluation research organization.
The investigation was reportedly triggered in part by OpenAI's July disclosure that its agents had breached Hugging Face during cybersecurity testing. At a New York City Council hearing on October 6, representatives from OpenAI, Anthropic, Meta, and Google all declined to guarantee that their AI agents would always follow safety guardrails, saying that eliminating risk entirely is not possible.
UK Regains Access to Frontier AI Models After 17-Day Freeze
The UK's AI Security Institute (formerly the AI Safety Institute) regained access to pre-release frontier models from Anthropic, Google DeepMind, and OpenAI on October 11, ending a tense 17-day standoff with the White House.
The freeze began on September 24 when the White House Office of the National Cyber Director ordered the companies to withhold new models from UK testers pending a US national security review. The reversal came after intense lobbying from UK AI minister Kanishka Narayan.
The episode highlights growing geopolitical tensions around AI safety governance. The UK institute has evaluated more than 20 frontier models to date, and the disruption raised questions about whether international AI safety cooperation can survive national security pressures. During the freeze, Anthropic reportedly released Claude Mythos 5.1 without giving the UK institute advance access.
Microsoft Transforms Copilot Into an OS for Work
Microsoft's redesigned Copilot, announced September 25 and now rolling out in preview, restructures the assistant around three layers: Home, Code, and Autopilot.
Code lets workers build apps and workflows from natural language, powered by GitHub Copilot technology. Autopilot (previously codenamed Scout) is a persistent agent that works autonomously with its own tenant identity, memory, and cloud workspace — CEO Satya Nadella called the new Copilot a new OS for work.
The Today command center enters private preview in October inside Copilot and will expand to Outlook and Teams. Cost tracking for Code and Managed Runtime is expanding through Agent 365, with Copilot Studio agent support also planned for October. General availability dates have not been announced, and Microsoft Ignite (November 17–20) is expected to bring further details.