On mvanemmerik.com, Nov 16: NFL Week 10 chalk talk: Detroit dazzles, Cincinnati takes the North
vanemmerik.aiClaude, ChatGPT and Grok news, twice a day

Thursday, November 5, 2026

OpenAI releases GPT-6 Cyber, a model it rates Critical for cyber risk, to vetted security teams

OpenAI's GPT-6 Cyber is going to approved Daybreak Red customers first, alongside its Images 2.6 GIF update, the Pierre deal and a pledge that its data centers will cover their own power costs. Anthropic launched Personal Finance, a read-only Plaid link for US Claude subscribers, plus new Enterprise security controls, a bigger Claude Marketplace and a Claude-assisted protein structure; it also reportedly licensed AP election data. Cat Wu joined OpenAI from Claude Code, and xAI was quiet.

Thursday, November 5, 2026

OpenAIModel

OpenAI releases GPT-6 Cyber to approved Daybreak Red security teams, the model behind its 948 Linux exploits

OpenAI launched GPT-6 Cyber on Nov. 5, a cybersecurity model available first to approved customers in Daybreak Red, its program for authorized exploit development and red teaming, through the API and Codex; it says all Red customers will get access over the coming weeks. OpenAI says the model found 81 previously unknown high-severity Firefox flaws, built the 948 Linux kernel exploits it described a day earlier, and beats GPT-6 Astra on most cyber tasks; it also rated the model "Critical" for cybersecurity under its Preparedness Framework, a step up from "High" for GPT-5.6 Cyber, and added stricter identity checks and misalignment monitoring. Pricing was not given, and the results have not been independently checked.

Why it matters: OpenAI now places one of its own models in the top cyber risk level of its framework, and is shipping it only to vetted defenders.

OpenAIOpenAI APICodexHappened Nov 5, 2026 · Sources: CyberScoop · The Next Web · Fortune

ClaudeLaunch

Claude adds Personal Finance in beta, linking US bank and card accounts read-only through Plaid

Anthropic began rolling out Personal Finance on Nov. 5 as a beta for Pro, Max and Ultra subscribers in the US. Users sign in to their bank through Plaid, and Claude gets read-only access to each linked account's name, type, balances, credit limits and up to two years of transactions, so it can track spending, flag fees and subscriptions, and build charts or budgets; it cannot move or invest money and is not meant to give investment advice. Anthropic says chats in the Personal Finance hub are not used for training, and accounts can be unlinked at any time.

Why it matters: It brings Claude into a sensitive everyday task where ChatGPT and banking apps already compete.

Try it: On a paid US plan, open Personal Finance from Claude's sidebar and connect an account through Plaid.

ClaudeHappened Nov 5, 2026 · Sources: Claude: Introducing Personal Finance in Claude · MacRumors

ClaudeFeature

Claude Enterprise gets a security center, inference hooks for DLP tools and a parent-organization admin console

Anthropic released a set of admin features for Claude Enterprise on Nov. 5. Inference hooks, now generally available, send prompts and tool calls to a company's own security tools before they run; in beta are a security center that compares an organization's chat, Cowork and Claude Code settings with Anthropic's recommended setup, centralized management at admin.claude.com for companies with several Claude organizations, one-call sign-out of a member, a self-serve IP allowlist, and blocking personal accounts in Claude Desktop on managed computers. A separate beta lets Enterprise and Team admins run their EDR sensor inside the Cowork virtual machine on macOS and Windows, in detection mode only.

Why it matters: Controls like these are often what security teams require before approving an AI tool company-wide.

ClaudeClaude CodeAnthropicHappened Nov 5, 2026 · Sources: Claude: More visibility and control over Claude Enterprise for security teams · Claude release notes

ClaudeBusiness

Claude Marketplace passes 25 partner products, adding Datadog, Runway, Asana, Zoom and five security vendors

Anthropic said on Nov. 4 that Claude Marketplace now lists more than 25 partner agents and products, with new additions including Chainguard, Checkmarx, Cyera, Rubrik and Tanium for security, Datadog and Netlify for developer infrastructure, AirOps, Runway and Webflow for creative work, and Asana, Glean, monday and Zoom for productivity. Eligible enterprise customers can apply part of their committed Anthropic spend to these tools.

Why it matters: Letting customers spend their Claude commitment on partner tools makes Anthropic's contracts stickier for large buyers.

AnthropicClaudeHappened Nov 4, 2026 · Sources: Claude: New agents and products on Claude Marketplace

ClaudeResearch

Anthropic scientists use Claude to help solve the first 3D structure of a bacterial vault particle

Anthropic's science team said on Nov. 4 that Claude, working under researchers' direction, searched about 15 billion predicted proteins, designed DNA constructs, suggested purification steps and ran cryo-EM processing, helping produce the first structure of a vault particle from prokaryotes, built from 96 copies of a phage protein. Anthropic says isolation time fell from about three days to 90 minutes, lists mistakes Claude made along the way, and notes the vault's function is still unknown. The work is in a bioRxiv preprint and has not been peer reviewed.

Why it matters: It is a concrete, documented example of an AI model doing hands-on lab science work, errors included.

AnthropicClaudeHappened Nov 4, 2026 · Sources: Anthropic: Investigating vault particles with Claude

ClaudeBusiness

Anthropic reportedly signed its first publisher deal, licensing AP election data for Claude before the midtermsReported

Adweek reports, citing two people familiar with the matter, that Anthropic signed a content and data licensing deal with the Associated Press just before the Nov. 3 midterms, and that Claude answered Election Night polling questions with AP-attributed information. Terms and length were not disclosed, and neither Anthropic nor the AP has confirmed the deal on the record.

Why it matters: Anthropic had avoided the publisher licensing deals OpenAI has signed for years, so this would mark a change in approach.

AnthropicClaudeHappened Nov 4, 2026 · Sources: Adweek

OpenAIPolicy

OpenAI pledges that future data center campuses it owns will pay their full power and grid costs

OpenAI published a set of voluntary commitments on Nov. 5 for future compute campuses it owns and operates: paying the full cost of the electricity they use and the grid upgrades they need, a local agreement and point of contact for each site, and annual public progress reports. Reports say it also pledged not to require NDAs from public officials and to publish water and energy use. A companion Compute Campus Policy Blueprint asks lawmakers to set similar rules for the whole industry.

Why it matters: Local opposition over power bills and secrecy has become a major hurdle for new AI data centers.

OpenAIHappened Nov 5, 2026 · Sources: The Chenab Times · Runtime Wire

OpenAILaunch

ChatGPT Images 2.6 adds animated GIFs and saved reference images, with two new API models, Flare and Sunburst

OpenAI released Images 2.6 on Nov. 4, rolling it out to all ChatGPT and Codex users on web, desktop and mobile. It can turn a prompt or photo into an animated GIF with adjustable speed and editable frames, save reference images of people, pets or characters for reuse, and adds image markup and SVG export. Developers get gpt-image-2.6-flare, a fast model OpenAI says can finish an image in about six seconds, and gpt-image-2.6-sunburst for higher quality and consistency across edits; OpenAI says API generation is about 50% faster.

Why it matters: It is the first time ChatGPT can make a GIF in one step, and the API split gives developers a clear speed-versus-quality choice.

Try it: In ChatGPT, ask for an animated GIF from a prompt or an uploaded image, then adjust the speed or edit single frames.

ChatGPTCodexOpenAI APIHappened Nov 4, 2026 · Sources: OpenAI Developer Community: Introducing ChatGPT Images 2.6 · The Next Web

OpenAISafety

OpenAI says an agent pipeline on an unreleased cyber model produced 948 working Linux root exploits for about $1.2 million

OpenAI published results from a mid-October run in which agents on an upcoming cyber-focused model searched the Linux kernel for bugs, reproduced 1,724 of 2,308 candidate findings, wrote 948 working local privilege-escalation exploits for a common distribution, and drafted patches. The run cost about $1.21 million in API credits, roughly $1,279 per exploit, and security researchers review findings before they go to kernel maintainers. OpenAI has not named the distribution or said how many bugs have been reported or fixed so far, and the results have not been independently checked.

Why it matters: It shows how quickly AI can now find and weaponize kernel bugs, which raises the pressure on patching to keep up.

OpenAIHappened Nov 4, 2026 · Sources: The Hacker News

OpenAIBusiness

OpenAI agrees to buy Pierre, maker of the Code Storage Git service and the Trees and Diffs libraries Codex already uses

OpenAI said on Nov. 4 that it will acquire Pierre Computer Company, taking on its team, its Code Storage hosting service and its open-source Trees and Diffs libraries, which Codex already uses for file trees and code diffs. Terms were not disclosed, the open-source projects stay public, and the deal still needs regulatory approval. It follows OpenAI's write-up of Pando, the Git-compatible backend it built for its own monorepo, which it says cut median push latency by about 80%.

Why it matters: It points to Codex growing from a coding agent into a place where code is also stored and reviewed, closer to GitHub's territory.

CodexOpenAIHappened Nov 4, 2026 · Sources: Investing.com · DevelopersIO: OpenAI's Pando

OpenAIBusiness

Update: Former Claude Code product head Cat Wu joins OpenAI to lead product management for Core ProductsReported

Days after leaving Anthropic, Cat Wu said on X that she is joining OpenAI to run product management for its Core Products group; OpenAI president Greg Brockman and Core Products lead Thibault Sottiaux welcomed her publicly. OpenAI has not published its own announcement, and her start date and reporting line were not stated.

Why it matters: A rival hiring the person who led Claude Code's product work is a notable talent move in the coding-agent race.

OpenAIClaude CodeHappened Nov 4, 2026 · Sources: Unite.AI

ClaudeBusiness

Google backs the financing of TeraWulf's $19 billion Kentucky data center for AnthropicReported

TeraWulf said on Nov. 4 that Google will provide a credit backstop for the financing of its Hawesville, Kentucky campus, built under a 20-year lease with Anthropic signed in July that TeraWulf expects to bring about $19 billion in revenue. The site is planned for about 401 megawatts of IT load, with first capacity in the second half of 2027. The size of Google's backstop was not disclosed, and Google and Anthropic did not comment in the Reuters report.

Why it matters: Google's credit makes it easier to fund the computing Anthropic has committed to as it heads toward its IPO.

AnthropicHappened Nov 4, 2026 · Sources: Reuters via 93.3 The Drive

OpenAIFeature

ChatGPT Dots will reportedly pick their own model for each task, adding GPT-6.1 Terra alongside GPT-6 AstraReported

OpenAI's Thibault Sottiaux said on X that the company is widening the models behind Dots: GPT-6 Astra stays, GPT-6.1 Terra is being added, and a Dot will choose the model for each job and may switch partway through. Reports say Terra should make Dots faster and help bring them to more users, but OpenAI has not posted release notes, a rollout date or whether users can override the choice.

Why it matters: A cheaper model in the mix could ease the cost of always-on agents, which has already squeezed usage on rival Grok Bot.

ChatGPT DotsChatGPTHappened Nov 3, 2026 · Sources: Tbreak · Dataconomy