Week in AI · Nov 2 – Nov 8, 2026
The week in Claude, ChatGPT and Grok
Money and security dominated the week. Anthropic's amended IPO filing showed $17 billion in third-quarter revenue alongside a $567 million operating loss, and reports had investors pushing back on its valuation. OpenAI released GPT-6 Cyber to vetted security teams, shipped ChatGPT Images 2.6 and reportedly banked $11 billion at a $1.25 trillion valuation, while Reuters reported years of internal security warnings. xAI was quieter, with Grok Bot phone calls reportedly reaching all US users.
Scorecard
Each company gets the same treatment: what it shipped or announced this week, from its own sources and reputable reporting.
Claude & Anthropic
24 stories
Filed an amended S-1 that warned future models may soon accelerate AI research, and shipped Personal Finance (beta), an Enterprise security center and an API IP allowlist. On the downside, Claude Code product head Cat Wu left for OpenAI, investors reportedly balked at a $1.5 trillion valuation, and researchers disclosed a Claude Dispatch extension flaw.
- Anthropic reportedly pays $3.4 billion for 1.7 GW of gas-fired generators to power its data centers (reported)
- Anthropic study of 207,049 research sessions finds scientists mostly use Claude for data analysis and writing
- Anthropic details how Claude 'dreams' between sessions to tidy agent memory, lifting a wiki test from 13% to 48%
- Ramp card data shows business AI spending flattening even as token use keeps rising, weeks before Anthropic's IPO (reported)
- Researchers say a browser extension could use Claude Dispatch to run code on a linked computer, a flaw reported in May (reported)
- Claude Enterprise adds project tags in beta to track spend by charge code or client-matter number
- Update: Investors reportedly push Anthropic to lower its $1.5 trillion IPO valuation after a market selloff (reported)
- METR says Anthropic's model progress has roughly doubled in pace, most likely because AI is speeding up its research (reported)
- 16 more
ChatGPT & OpenAI
23 stories
Released GPT-6 Cyber to approved Daybreak Red teams, launched ChatGPT Images 2.6, agreed to buy Pierre for Codex and reportedly closed an $11 billion tranche. It also faced a Reuters report that staff had warned for years about lax internal security, and responded with Hacktron and Cathedral security partnerships.
- OpenAI partners with Hacktron, the security firm whose July report chained a forum bug into OpenAI's internal code
- OpenAI and Ligo publish BioBasicsBench, 100 routine wet-lab tasks where GPT-6 Astra scores 66.7%
- Update: OpenAI reportedly banks the first $11 billion of its new round at a $1.25 trillion valuation (reported)
- Codex CLI 0.167 adds an agent command center that marks unread results and a Vertex AI setup wizard
- Greg Brockman says OpenAI could be ready for an IPO in 2027 as enterprise becomes most of its revenue (reported)
- OpenAI releases GPT-6 Cyber to approved Daybreak Red security teams, the model behind its 948 Linux exploits
- ChatGPT Images 2.6 adds animated GIFs and saved reference images, with two new API models, Flare and Sunburst
- OpenAI says an agent pipeline on an unreleased cyber model produced 948 working Linux root exploits for about $1.2 million
- 15 more
Grok & xAI
3 stories
A quieter week: xAI showed Data Bot, an internal Grok Bot team agent, retired two Grok Imagine image API models in favor of a cheaper setting, and Grok Bot phone calling reportedly reached all users for US numbers. Users kept complaining about Grok Bot weekly limits after Elon Musk said it cost too much to serve.
- xAI retires its grok-imagine-image-quality and -pro API models, serving them with a cheaper Grok Imagine Image 2.0 setting
- Update: Grok Bot's phone calling reportedly reaches all users, for US numbers only (reported)
- xAI shows Data Bot, an internal Grok Bot team bot that queries its warehouse and posts reports to Slack
Pick of the week
hunt for exploitable bugs in your own code and confirm the fixes work
Best fit this week: OpenAI. OpenAI's GPT-6 Cyber, released Nov. 5, is built for finding vulnerabilities, reproducing them in an authorized test environment and checking that patches hold, which matches this job most directly. Access is limited to approved Daybreak Red security teams, so most developers will still reach for a general coding agent.
Runner-up: Claude. Based on this week's announcements and current features, not a benchmark; every tool is judged the same way.
Models and prices
- OpenAI ChatGPT Images 2.6 released Nov 4, 2026
- OpenAI GPT-6 Cyber released Nov 5, 2026