Scoop
AI tool news · rumor vs. reality

The AI Wire ●

Rumors tracked. Announcements verified. Updated daily.

2026-11-09

RELEASED

OpenAI puts ChatGPT Enterprise on AWS Marketplace: buy and bill it through your existing AWS account

OpenAI announced on November 9 that ChatGPT Enterprise is now available through AWS Marketplace, so eligible organisations can discover, buy and manage it through their existing AWS purchasing and billing relationships instead of contracting with OpenAI directly. Eligible spending can also count toward a customer's AWS commitments (EDP or PPA), subject to their agreement. The listing names OpenAI as the seller and carries no standard price — customers request a private offer with a custom quote.

OpenAI says marketplace customers get the same models, features and functionality as direct customers, with new features released at the same time; the package combines Chat, ChatGPT Work, Codex and Dots. Eligible customers can also access ChatGPT for Healthcare and ChatGPT for Financial Services through the Marketplace, confirmed through an OpenAI or AWS account team.

One boundary worth noting: procurement is not data residency. The listing states that business data is encrypted at rest and in transit and is not used to train OpenAI models by default, but buying through AWS Marketplace does not guarantee that all customer data remains on AWS infrastructure. The launch changes how organisations contract and pay for ChatGPT Enterprise, not what the product does.

RELEASED

OpenAI swaps ChatGPT Ads bidding to Target CPA and Target CPC from today — Max Bid retires January 1

From today, November 9, new conversion campaigns in ChatGPT Ads bid with Target CPA instead of Max Bid, and new click campaigns can use Target CPC instead of Max CPC, OpenAI said in a Help Center update dated November 5. Target CPA and Target CPC are average cost goals: the system may bid above or below the target in any single auction to hit the average.

Existing Max Bid and Max CPC campaigns keep serving on their current strategy until January 1, 2027, when OpenAI moves the remaining campaigns to the matching target automatically. OpenAI told advertisers to set targets from their recent CPA or CPC, and to fix Event Quality Score warnings on their conversion events before the switch.

For advertisers this is a forced migration, not an optional refresh: targets set far from recent actual costs risk stalled delivery after the January changeover. The change lands in an ads business that has grown fast — ChatGPT Ads had reached a $1 billion annualised revenue run rate by August 31, in under 200 days, according to the company.

ANNOUNCED

NVIDIA unveils GR00T 2, a robot model that learns manipulation from human video, at CoRL

NVIDIA introduced GR00T 2 at the Conference on Robot Learning (CoRL) in Austin on November 9, calling it an open frontier world action model built on its Cosmos world models. The 2-billion-parameter model predicts future video and robot actions together, and was trained on diverse robot and human data, including about 402,730 hours of first-person human video.

NVIDIA reports GR00T 2 completed 40 of 50 evaluated tasks at least once with no task-specific demonstrations, reached 71% average success with one demonstration per task, and 95% with five. One pretrained checkpoint covers parallel-gripper arms, dexterous hands and a humanoid, and NVIDIA is releasing 100,000 hours of the human egocentric data openly under a CC BY 4.0 licence.

Alongside the model, NVIDIA shipped Isaac Lab 3.0, its open-source robot-learning framework, which lets agents set up training environments and apply a new reinforcement post-training library to foundation models like GR00T. The GR00T 2 code is listed as arriving in December, and no licence or release date for the model weights has been announced yet — Foxconn, Siemens and Nexuni are testing the model, not deploying it.

RELEASED

Microsoft Copilot can now create, update and delete data in connected apps — connector write actions are generally available

Microsoft has rolled out connector write actions as generally available across Copilot Chat and Cowork, Windows Report reported on November 9. Until now Copilot's connectors were mostly read-only — find information, summarise conversations. With write actions, Copilot can also create, update and delete records in connected third-party applications, turning a conversation into completed work in the app where the data lives.

Microsoft says Copilot does not automatically execute every requested change, and that write actions will also arrive in Copilot across the Office apps soon. No list of which connectors support writes at launch was published in the reporting.

This is a practical shift for the app everybody uses at work: a Copilot that can only read is a search box with better manners, while one that can write back into Salesforce, ServiceNow and the rest is an agent. The control question — which writes need a human confirm — is the part to watch as the Office-app rollout lands.

RELEASED

Meta puts all three Muse models on rival Google Cloud: Ember, Spark 1.3 and Wonder live in Model Garden

Meta has made its three Muse reasoning models — Muse Ember, Muse Spark 1.3 and Muse Wonder — available on Google Cloud, as managed APIs in Model Garden on Google's Agent Platform. Mark Zuckerberg announced the move on November 9, the same day Ember launched; Google CEO Sundar Pichai replied publicly welcoming the models less than an hour later.

The models are served through an OpenAI-compatible Responses API on Google Cloud's global endpoint, using the model IDs muse-ember-1.0, muse-spark-1.3 and muse-wonder-1.0. Google Cloud now hosts models from Meta alongside its own Gemini models and offerings from OpenAI and Anthropic. The larger Muse Wonder Full was not included — it remains available only through a Muse Code subscription.

Neither company has published what the Muse models cost on Google Cloud. On Meta's own Model API, Ember is $0.10 per million input tokens, Spark 1.3 is $1.25/$4.25 and Wonder is $4/$20, but those are Meta's prices, not Google Cloud's.

RELEASED

Meta releases Muse Ember, its lightest model yet, at $0.10 per million input tokens for agent volume work

Meta Superintelligence Labs released Muse Ember on Monday, November 9, the lab's lightest and most cost-efficient model yet and the fourth model under the Muse brand, after Spark in April, its open-source twin Glimmer in August, and the Wonder frontier model in October. Meta says Ember is designed for complex reasoning execution and the work agents repeat all day: tool calls, coding subtasks, routing, extraction and long document parsing. RuntimeWire, citing Mark Zuckerberg's announcement post on X, says Ember is available on the Meta Model API and in Muse Code.

Price is the headline: $0.10 per million input tokens and $0.50 per million output tokens. Meta pitched Ember on cost per task, using Artificial Analysis's composite benchmark score to present it as cheaper and smarter than Anthropic's Haiku and OpenAI's Luna, and cost-competitive with open-source rivals DeepSeek V4.1 Flash and GLM 5.3 Flash. The intended pairing is a split of labour — in Meta's words, 'Muse Wonder plans, Muse Ember handles the volume.'

Ember is also multimodal: Meta says agents can read screenshots, charts and scanned documents, or work from voice memos, meeting recordings and video, without a separate model. No independent benchmark report for Ember had appeared at launch; the cost-per-task comparisons are Meta's own, built on the Artificial Analysis composite.

RELEASED

Meta brings its Muse AI agent to Apple Watch and Wear OS

Meta has launched its Muse AI agent on Apple Watch and Wear OS, announced Monday, November 9. On a smartwatch, users can reach Muse by voice call, text message or dictation directly from their wrist, as a companion to the Muse app on their phone.

The watch launch continues a fast device rollout: Muse first arrived on the App Store in September, got a dedicated iPad app in October, and added Spaces for organizing tasks on November 6. Meta says Muse passed 10 million U.S. downloads in its first 40 days, per Sensor Tower data cited in coverage.

Meta has not detailed which Muse features work fully on the watch versus the phone, or published regional pricing and availability beyond the initial rollout. A dedicated Muse Charm device, teased at Meta Connect in September, is still expected before the end of 2026.

ANNOUNCED

Graphite: Claude passed Gemini in monthly chatbot sessions in June — 5.44 billion to 3.72 billion in September, with ChatGPT far ahead

Claude passed Gemini in monthly chatbot sessions in June and has stayed ahead every month since, according to a report published on November 9 by Graphite, a research and growth marketing firm. Graphite estimates Claude had 5.44 billion worldwide web visits and mobile app sessions in September, against 3.72 billion for Gemini. ChatGPT remains far ahead of both, with an estimated 61.35 billion sessions in September — 81.2 percent of the total across the six chatbots in the report.

Graphite built the count from Similarweb estimates. A session is one stretch of use, so the figures measure how much people use each chatbot, not how many people use it. Graphite puts Claude's session growth at 1,901 percent over the year, and says the surge appears to have come from the launch of Claude Cowork, the Opus 4.6 and Sonnet 4.6 model releases, and Anthropic's widely publicised standoff with the Pentagon.

The count has clear limits: it covers web and app sessions only, leaves out API and developer use, and Graphite says it has no reliable data source for desktop apps, so desktop use is understated. The report landed as Anthropic prepares its IPO: Bloomberg reported on November 9 that Anthropic can begin formally marketing the offering to investors as early as November 10, while its amended S-1 from November 2 still leaves the price range blank.

ANNOUNCED

Google announces TPU 8tx for confidential inference and TPU 8i Max for massive context windows — four TPU systems in one year

Google announced on November 9 that it is expanding its eighth-generation TPU portfolio with two new systems: TPU 8tx, built for secure and confidential inference, and TPU 8i Max, built for massive context windows and the highest compute density per rack. They join TPU 8t for large-scale training and TPU 8i for real-time reasoning, both introduced at Google Cloud Next in April — the first time Google has introduced four TPU systems in a single year.

TPU 8tx uses the same silicon as TPU 8t, paired with 5th Generation AMD EPYC hosts, and builds security into the hardware so sensitive data and model weights stay protected and isolated even while in active use; Google claims up to 2x lower latency than TPU 8t and says it also suits small-scale training and private fine-tuning for regulated enterprises and governments. TPU 8i Max puts two TPU 8i chips on a single package, doubling compute, SRAM and HBM bandwidth over TPU 8i, aimed at workloads where dozens of agents work across a whole codebase or hours of video at once.

Google's post gives no price and no general-availability date for either system; Bloomberg reporting cited by Beri says the systems arrive next year. Google says TPUs now have nearly 100 customers, including frontier labs, trading firms, retailers and universities. All performance figures are Google's own.

ANNOUNCED

Google's own help pages now say Gemini's human reviewers can see excerpts of your connected Gmail and Drive — the 2023 'not seen by human reviewers' promise now carries an October 16 caveat

A write-up documenting Gemini's data handling topped Hacker News this week, and its core evidence is Google's own wording. Google's Connected Apps help page says "a subset of Connected Apps data processed by Gemini is reviewed by human reviewers (including trained service providers)" — the summaries, excerpts and inferences Gemini draws from your emails and files, and, in Google's words, when an email or file is short or especially relevant "the summary may just be the email or file."

That is a quiet reversal of the promise Google made in September 2023, when it first let its chatbot read Gmail, Docs and Drive: that content "is not seen by human reviewers." The 2023 post still says that, but now carries an October 16, 2026 note qualifying it for newer features. Google's privacy hub adds that reviewed chats are kept for up to three years, disconnected from the account, and are not removed when a user deletes their Gemini activity.

The contractor-review clause applies to personal Google accounts using Connected Apps and Personal Intelligence; work and school Workspace accounts are treated differently, with Google saying their content is not human reviewed or used for training outside their domain without permission. The practical takeaway for personal users: if you do not want a reviewer to potentially see it, disconnect Gmail and Drive in Gemini's settings — Google's own page ends with that warning.

RUMOR Rumor still unresolved

Rumor: a Google internal checkpoint called 'Gemini 4 Krypton' is said to beat Carbon — no scores, no catalog entry and no word from Google

A post by Lyra on November 9 describes a new Google internal checkpoint called Krypton, saying its early results are impressive and 'even better than Carbon.' That is the whole of the traceable claim: no published tasks, scores or reproducible testing material accompany it, and the word 'internal checkpoint' is the key phrase — nothing in the post describes a product Google has announced.

A check on November 16 found no Krypton entry in Google's public model catalog, and searches for a formal API ID return no public active routes. Krypton has no confirmed API ID, price or release date. The November and December launch windows circulating in discussion are a user's hope and a prediction in a post title, respectively — neither establishes a release window, so this stays a rumor, unresolved.

ANNOUNCED

Fireworks AI expands its breach disclosure: a second intrusion path reached billing data and, for a limited set of accounts, secrets, datasets and partial model weights

Fireworks AI used a November 9 update to widen its disclosure of an October security incident. The company first found unauthorized use of internal credentials on October 6; the update adds a second path — a third-party cloud provider hosting Fireworks R&D machines was breached on October 23, and the next day the intruder used it to reach Fireworks development machines and run an unauthorized workload. Fireworks says it detected that activity the same day and has seen no further successful access on its customer platform since October 25.

The expanded impact list covers account and billing information — billing and shipping addresses, invoice and payment history, account balances and credit usage — and, for a limited set of accounts, secrets, datasets and a subset of incomplete model weights. Payment card numbers are not stored by Fireworks and were not affected, and the company says it has detected no impact to Inference, the Training SDK or Nexus, which are zero data retention by default, though the investigation is ongoing.

Fireworks has disabled self-serve dataset uploads and downloads while the investigation continues, brought in Palo Alto Networks Unit 42, and notified law enforcement. It links the activity to a financially motivated group engaged in LLMjacking — reselling unauthorized access to large language models.

RELEASED

Anthropic releases Claude Fable 5.2 and Mythos 5.2: up to 30% faster, $8/$40 per million tokens, top score on the Artificial Analysis index

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday, November 9 — two configurations of the same underlying model, and the company's most capable generally available models yet. Fable 5.2 carries safeguards that block most cybersecurity and biology requests, which is what lets Anthropic offer it broadly; Mythos 5.2 drops those blocks and is limited to organisations vetted through Anthropic's Cyber Verification Program or Life Sciences Verification Program. Fable is available now in Claude's paid plans, Claude Code, the Claude Platform (model ID claude-fable-5-2), and on AWS, Google Cloud and Microsoft Azure; Cognition made it available in Devin Desktop and Devin CLI on launch day.

The headline numbers: output up to 30 percent faster than the 5.1 generation, clearer communication, and a price cut — $8 per million input tokens and $40 per million output, down from $10/$50, with an optional fast mode at $16/$80 for up to 2.5x generation speed. Platform documentation lists a 1M-token context window and 128K maximum output. Anthropic says Fable 5.2 took the top spot on the Artificial Analysis Intelligence Index v5 at 65, and reports 49.1 percent on FrontierCode 2.0 at Xhigh effort, up from 32.4 percent for Fable 5.1. Paid-plan subscribers get extra Fable usage resets during launch week.

Anthropic still recommends most users start with Claude Opus 5.5 for everyday coding and knowledge work, moving to Fable or Mythos for the most ambitious, open-ended technical work. The launch is Anthropic's first frontier release since CEO Dario Amodei called in September for a slowdown in AI development, and it lands days ahead of the company's expected IPO marketing. Launch-day demos spread fast: Matthew Berman showed Fable 5.2 building a playable single-level Gears of War clone in Unreal Engine from a three-line prompt over a few days of unsupervised work. All benchmark and speed figures are Anthropic's own.

2026-11-08

RELEASED

SpaceXAI widens Grok Bot access: SuperGrok, Cursor Pro and every Cursor Teams plan now get the agent

Grok Bot, SpaceXAI's agent that works inside other software on its own persistent cloud computer, is now much easier to get. Engadget reported on November 8 that the company has expanded it to SuperGrok subscribers, Cursor Pro and all Cursor Teams plans — adding the agent to cheaper tiers while keeping it on the more expensive ones. X Premium+ subscribers can also link their accounts for access.

The plan map is sprawling: Cursor Pro costs $20 a month, Pro+ $60 and Ultra $200, while Teams runs $40 to $120 per user per month; on the Grok side, SuperGrok is $30 a month, SuperGrok Plus $100, and X Premium+ $40 in the US. Included Grok Bot usage resets weekly, extra usage can run through an account's shared on-demand spending, and there is a usage-based free trial with a seven-day window.

Grok Bot is still in beta. Each Bot gets a browser, a file system and a terminal and can keep working after your own computer is closed, but Engadget notes that websites can still block it with CAPTCHAs and login steps, and early users say some jobs need more supervision than the 'go away and do it' pitch suggests.

ANNOUNCED

Nvidia's board opens an independent investigation into reports CEO Jensen Huang was targeted by a foreign intelligence effort

Nvidia's board of directors said on Sunday, November 8, that it is opening an independent investigation into reports that chief executive Jensen Huang was the target of a foreign intelligence effort. A committee of independent directors, assisted by outside counsel, will oversee the review. The board said it has 'no reason to believe' that either national security or Nvidia secrets were compromised, and that Huang fully supports the review, has committed to cooperating, and will remain CEO while it runs.

The announcement followed New York Times reporting that U.S. intelligence agencies determined Alice Huang, a Chinese citizen and former executive at Singapore-based data-centre company Megaspeed, was gathering information for Beijing. According to the Times, Huang acknowledged to the board that he had been intimate with her, and apologised in a message to staff and family. Megaspeed has bought more than $4.8 billion worth of Nvidia chips over the past two years, the Times reported; the two Huangs are not related.

Nvidia's chips power much of the AI build-out and sit at the centre of U.S. export restrictions on China, which is why a report of this kind reaches the board and Congress: three U.S. House committee chairmen have asked Huang to attend a closed-door briefing on November 18. China's foreign ministry said on Monday that the spy story was unsubstantiated, Reuters reported. Nvidia gave no timetable for the investigation.