Scoop
AI tool news · rumor vs. reality

The AI Wire ●

Rumors tracked. Announcements verified. Updated daily.

2026-11-05

RELEASED

Anthropic rolls out Personal Finance in Claude: link your bank via Plaid, read-only, in beta for US paid plans

Anthropic announced on November 5 that Personal Finance is rolling out in beta this week inside Claude, for Pro, Max and Ultra subscribers in the US. Users link bank and card accounts through Plaid — which connects to more than 12,000 financial institutions — and can then ask Claude about balances, spending and bills in plain language.

Access is read-only: Anthropic says Personal Finance cannot move, manage or invest money, and is not designed to give investment advice. What Claude can see is bounded — the bank's name, account name and type, the last four digits, balances, credit limits, and up to two years of transactions — and Anthropic says the hub carries no ads or sponsored recommendations and that its conversations are not used to train its models. Suggested uses include spotting unexpected fees, double charges and bills that have crept up over the year.

The launch confirms the Wire's October 18 rumor: TestingCatalog first spotted a 'Money' tab in the Claude iOS app on September 14, then Plaid-linked Personal Finance screens in October. Claude arrives on the same Plaid connector ChatGPT Finances used in May, putting a second major assistant on live bank data — with the same caveat Anthropic prints itself: Claude can make mistakes, so check important figures with your bank.

RELEASED

Anthropic ships a Claude Enterprise security package: inference hooks are generally available, a beta Security Center checks your setup

Anthropic's November 5 security update for Claude Enterprise is led by inference hooks, now generally available for Enterprise organizations. Prompt hooks send a user's prompt to the organization's own data-loss-prevention endpoint before Claude generates anything; pre-tool-call hooks check a proposed tool call before it runs. The verdict is binary — allow or deny, with a default five-second timeout — and the hooks work with endpoints from Netskope, Palo Alto Networks, Proofpoint and Zscaler, or a customer's own server.

The hooks apply across claude.ai on web, desktop and mobile, plus Cowork, Claude Code, Claude for Microsoft 365, Claude in Slack and Claude in Chrome — one setup inside Claude's infrastructure rather than on each employee device. They are not available for organizations using Claude through Amazon Bedrock or Google Cloud.

The rest of the package is still in beta: a Security Center that lines an organization's chat, Cowork and Claude Code settings up against Anthropic's recommended configuration, flags the gaps and exports an auditor report — guidance, Anthropic stresses, not a compliance certification; centralized management for parent organizations at admin.claude.com; and an endpoint-detection sensor that can run inside the local Cowork virtual machine, though cloud Cowork sessions, which run on Anthropic's servers, cannot host one.

ANNOUNCED

Claude Code was putting users' email addresses in web-request headers — fix shipped, gaps reported

A bug report filed on November 5 in Anthropic's claude-code GitHub repository (issue #107757) says Claude Code sent the signed-in account's email address in the User-Agent header of web requests the agent triggered, exposing the address to arbitrary third-party sites and their server logs. The reporter links the leak to receiving dozens of scam emails a day and argues the behaviour should be removed from the harness entirely, not made a setting.

An Anthropic contributor replied the same day that a built-in mod, cc-plugin-privacy, introduced in version 2.1.317, deterministically blocks the behaviour and is enabled by default. Follow-up reports say the coverage is incomplete: the mod does not govern network scripts the agent writes and runs itself, at least one user on version 2.1.318 reports a test header still passing, and a separate issue open since July (#81138) documents the same email address being injected into the model's system prompt.

An independent test of the published fix code, posted November 6, ran 23 leak shapes against it: the exact reported leak is blocked, but 8 other shapes passed through. The GitHub issue remains open, and Anthropic staff say they are also removing the account email address from the context Claude Code gives the model.

ANNOUNCED

Agility Robotics begins trading on Nasdaq as AGLT — the first US-listed pure-play humanoid robot maker

Agility Robotics began trading on Nasdaq under the ticker AGLT on November 5, after its merger with Churchill Capital Corp XI closed on November 4. Churchill shareholders approved the deal on November 2 with more than 84% of votes cast in favour. The combination, struck at a $2.5 billion pre-money valuation, brought in roughly $465 million in gross financing, and the shares opened at $11.26 against the $10 SPAC price before a volatile first two days.

Agility describes itself as the first US-listed company built entirely around humanoid robots with active commercial deployments. Its two-legged Digit robots are working with GXO, Schaeffler and Toyota Motor Manufacturing Canada; the company says Digit has moved more than 100,000 totes on a paid GXO contract and logged 65,000 hours across commercial operations, and it discloses more than $300 million in conditional, multiyear orders for the next-generation Digit 5.

The caveat is the gap between the listing and the business underneath it. Digit 5 reaches early customers only in the first half of 2027, with general availability targeted for the end of that year, and those orders depend on contractual milestones. Agility's 2025 sales were $1.8 million, and its accumulated deficit stood at $675 million by mid-2026 — the public market is funding the robot race well ahead of the revenue.

2026-11-04

RELEASED

Tencent releases the full Hy4, with vision and computer use — and says it deliberately starved the model of offensive cyber data

Tencent released the full version of Hy4 on November 4, adding vision and computer-use capabilities and making the model available through Tencent Cloud and OpenRouter and inside CodeBuddy, WorkBuddy, Yuanbao and ima, Unite.AI reported. The August preview was open-sourced at 770 billion total parameters, 49 billion active and a 1M-token context window; the full release keeps the listed API prices unchanged at $0.834 per million input tokens and $2.501 per million output, with OpenRouter listing a 1,048,576-token context and image input. The weights and a comprehensive system card are planned for a later release, and a cheaper, faster Hy4 Flash is planned after that.

The release's most notable choice is what Tencent left out. Reuters reported that Hy4 was trained with limited cybersecurity data — defensive content only — to reduce its ability to aid cyberattacks: the company filtered out exploit development, reverse engineering and penetration-testing material, keeping defensive proof-of-concept generation and vulnerability patching, and says red-team audits by its Xuanwu and Keen security labs found no elevated risk on tasks like defeating binary obfuscation. Moonshot AI took a similar step with its newest model last month, as Chinese labs follow the US labs in curbing offensive capability.

Tencent reports gains over the preview in instruction following, long-context understanding and long-horizon tasks, and says hallucinations in internal search evaluations fell from 15.76% to 5.45%. It also concedes a gap remains to top proprietary models on research-level mathematics, long-horizon coding and advanced visual question answering. The benchmark claims are company-run and internal; the hosted model, however, is live today, while open weights remain a promise.

ANNOUNCED

Reuters: shopping agents are splitting retail — Amazon has blocked Meta's Muse, Walmart is partnering with it

AI agents that research and buy products on a shopper's behalf — like Meta's Muse — are posing a conundrum for the big retailers, Reuters reported on November 4: let agents shop your store and you may lose the lucrative advertising brands pay for placement, plus the customer data that fuels it. Morgan Stanley researchers estimated in July that such agents could account for 10% to 20% of US e-commerce spending by 2030.

The two giants have split. Amazon banned Muse from its website in September, saying the agent appeared to have accessed its store without authorisation, capturing and storing customer credentials; Meta declined to comment. Amazon told Reuters this week that third-party agents should operate openly and respect service providers' decisions about whether to participate.

Walmart went the other way, rolling out infrastructure on October 21 to enable third-party agentic commerce and officially partnering with Meta's tool. The split is the shopping-agent question in miniature: the agent works for the shopper, but the store it shops in is an advertising business — and the two models do not obviously fit together.

ANNOUNCED

OpenAI's agents found 948 distinct Linux root flaws in one run — each with a candidate patch, for $1.2M in API credits

OpenAI published research on November 4, 'A thousand fewer paths to root,' describing a mid-October run in which an autonomous pipeline built on an upcoming cyber model hunted flaws in the Linux kernel. Scout agents picked 5,034 places to audit — about 2.8 million lines, roughly 7% of the kernel — and the pipeline turned 2,308 candidate findings into 1,724 dynamically reproduced vulnerabilities.

After variant analysis grouped exploits tracing back to the same flaw, OpenAI counts 948 distinct local privilege-escalation vulnerabilities, each backed by a working exploit — counted only when it produced a root shell on a stock desktop system with standard protections on — and a candidate patch. The whole run cost $1,213,430 in API credits, with exploitation the most expensive stage at $775,929; OpenAI values the haul at $18.9 million to $66.3 million at past bug-bounty contest rates, and puts the model's outright incorrect analysis at 3.1%.

The bounds matter: 948 is a count of distinct flaws with candidate patches, not fixes already on anyone's machine, and the post names no CVE identifiers, kernel versions or subsystems. Ten days after the scan, OpenAI says 61 flaws had already been patched upstream, 43 had patches in staging and 299 were being reported and discussed. The work sits under Patch the Planet, part of OpenAI's Daybreak programme, with human review before anything reaches a maintainer.

RELEASED

OpenAI makes GPT-Live outbound phone calling generally available to all organizations at $0.05 a minute

OpenAI has made outbound calling for GPT-Live generally available to all organizations, so voice agents can now dial real phone numbers themselves instead of only answering inbound calls. The calls run over SIP trunks from providers such as Twilio, Telnyx and Ozonetel, which carry the call onto the public phone network while GPT-Live handles the conversation.

The voice layer is priced at $0.05 per minute, the same rate as GPT-Live-1 in the API, with the SIP provider's carrier charges billed separately. Calls are full-duplex — both sides can speak and listen at once — and support HD Voice and keypad tones, so an agent can navigate automated phone menus. Signalling runs over TLS and audio over SRTP, with a European endpoint for data-residency needs.

SIP joins WebRTC and WebSocket as a third way to connect GPT-Live to audio. GPT-Live-1 reached the API in September with SIP connectivity for telephony; the new step is outbound dialling through the Live API, opened to every organization rather than a limited set.

RELEASED

OpenAI launches ChatGPT Images 2.6: native GIFs, reusable reference images and two faster API models

OpenAI launched ChatGPT Images 2.6 on November 4, adding native animated GIF creation to ChatGPT and Codex. Users can turn a written prompt or an uploaded photo into a looping GIF, then change its speed, revise individual frames and share the result — a one-step route ChatGPT did not have before.

The release is available globally across all subscription tiers on web, desktop and mobile, and OpenAI says image volume now exceeds 4 billion images a week across ChatGPT, Codex and the API, up from more than 3 billion at the Images 2.5 launch on September 8. Images 2.6 also adds reusable reference images, aimed at keeping a subject consistent across successive edits.

For developers, OpenAI added two API models: GPT-Image-2.6 Flare, positioned for faster generation — RuntimeWire reports output in as little as six seconds and generation about 50% faster — and GPT-Image-2.6 Sunburst, positioned for higher-quality output and consistency across edits. The split mirrors the 2.5 lineup, separating fast draft work from premium production work.

ANNOUNCED

OpenAI to acquire Pierre Computer: Code Storage Git hosting comes to Codex

OpenAI announced on November 4 that it will acquire Pierre Computer Company, the small developer-tools startup behind Code Storage, a Git hosting service built for AI apps and coding agents. Code Storage will be integrated into Codex; Pierre's open-source projects will remain open source after the deal closes.

Codex users have already touched Pierre's work: its Trees library powers file browsing in the Codex app and its Diffs library powers change review. Code Storage is used by agent builders including Lovable, Bolt and Amp for code, memory files and documents. Pierre was founded by Jacob Thornton, a co-creator of Bootstrap, and Ian Ownbey, an early Twitter engineer, and raised a $22 million Series A in November 2025.

No price was disclosed, and the deal still needs regulatory approval; until it closes, the two companies remain independent. Axios, which broke the news, framed it as OpenAI going after Microsoft's GitHub — and it landed the same day Meta acquired Conductor, the parallel coding-agent startup, underlining that the coding race has moved from generating code to the infrastructure around it.

ANNOUNCED

Midterm backlash: Texas and Georgia elect governors who promised data-centre pauses, and 33 local limits pass or lead

Opponents of AI data centres were among the biggest winners of Tuesday's midterm elections, Reuters reported on November 4. In Texas — which GLJ Research estimates holds about a quarter of the US data-centre capacity due online by 2030 — Democrat Gina Hinojosa ousted Governor Greg Abbott after promising a moratorium on new data-centre construction, stricter water and power rules and an end to industry tax breaks; the Associated Press called the race at 11:17pm Eastern, and Hinojosa led 52.8% to 45.7% with about 93.5% of the expected vote counted. In Georgia, where CBRE says Atlanta is the country's most active data-centre construction market, Keisha Lance Bottoms won the governorship on a call for a pause on new builds.

The local votes went the same way. Voters in 10 states faced 41 data-centre measures, Ballotpedia counted — 21 of them in Ohio alone — and by Wednesday 33 measures restricting data centres had passed or were leading, according to Stateline's tracker. In Ohio, Democrat Amy Acton beat Vivek Ramaswamy on a conditional freeze of new approvals unless strict energy and water criteria are met. An NBC exit poll found 56% of voters said AI and data centres were a crucial or major factor in their congressional vote.

The results are still unofficial, and what the new governors can actually do depends on state law and their legislatures — the pledges also differ in strength, from an immediate moratorium in Texas to a temporary pause in Georgia and a conditional freeze in Ohio. But investors told Reuters the breadth of the Democratic wave points to a shifting public mood on AI and a possible slowdown of the construction boom that has been propping up US growth and stock markets. This follows the Wire's November 3 entry on the moratorium pledges that were on the ballot.

Update, midday November 4: the market has started pricing the backlash in. Nasdaq-100 futures fell 0.8% overnight, shares in AI cloud company CoreWeave dropped 7% and Oracle 4% in overnight trading, CNBC reported. The exposure is concentrated exactly where the votes landed: Wells Fargo puts 27% of the gigawatts of planned US data-centre capacity in Texas, 7% in Georgia and 5% each in Ohio and Pennsylvania — and OpenAI, Anthropic, Google, Meta and xAI all need vast new capacity in precisely those kinds of states.

ANNOUNCED

Microsoft slips Copilot usage-based billing rollout to December — GPT-6 Astra and Claude Fable 5.1 included

Microsoft has pushed the general rollout of usage-based billing (UBB) models in Copilot back by about a month. In a November 4 update to Message Center post MC1519122, the Frontier preview runs from early to mid-November, but general availability now begins in early December and completes by mid-December — previously it was slated for early to mid-November.

The models arriving under UBB are not minor: GPT-6 Astra and Claude Fable 5.1 will be generally available as usage-billed models in Copilot Chat and in Word, Excel and PowerPoint via the model selector. Microsoft also says users will be able to set the reasoning effort level when using a UBB model in Chat.

Separately, Microsoft's partner guidance confirms usage-based billing will be enabled by default starting December 1, 2026 for new Microsoft 365 Copilot Business licenses bought through CSP. For organisations budgeting Copilot, the billing model is changing on a schedule that has already moved once.

ANNOUNCED

Microsoft caps more Copilot Chat features for staff without a paid Copilot licence — and won't say which ones or by how much

Microsoft is extending its fair-use limits in Copilot Chat, the AI chat bundled at no extra charge into Microsoft 365 business plans, to more features for employees who do not have a paid Microsoft 365 Copilot licence. It told partners in its Partner Center on November 4, without saying which features are capped or at what thresholds. Until now the limits applied only to a few features, such as image generation.

The cuts took effect on November 2 for organisations with up to 2,000 seats, and reach larger organisations from November 9. Anyone who hits a limit sees a notice saying when the feature will be available again — an answer already being written is not cut off — and in most organisations the notice includes a button to ask the admin for the paid licence. The change covers Copilot Chat on the web, desktop and mobile.

Admins cannot look up or change the caps: Microsoft says it does not publish them because they may change, and that they shift during the day. Computerworld reports analysts warning that unpredictable limits may cut productivity and push staff toward AI tools the company has not approved.

ANNOUNCED

Meta has paid $1.2M in Muse bug bounties — including $300,000 for a pre-launch flaw that exposed conversations and tokens

Meta has paid roughly $1.2 million in bug bounties for its Muse personal AI agent, the company said in a blog post shared by David Singleton and reported by SecurityWeek on November 4. The programme ran privately from April and went public at Muse's September 8 launch; of more than 630 reports received, 64 were valid. Meta pays up to $300,000 per valid report — several times the top rewards it says other leading AI developers advertise — with prompt-injection attacks on a single user worth up to $130,000 and a fleetwide compromise up to $250,000.

The largest award, the full $300,000, was paid four months before launch, while Muse was still in early internal testing: a researcher demonstrated access to conversations and authentication tokens on the pre-release virtual machines Meta staff used to test the product. An authentication check had treated a header added by a local HTTPS proxy as proof a request came from a trusted source, and a system migration added a second proxy layer that let outside requests carry the trusted header too. Meta took the service offline, fixed the bug and stopped treating the local machine as trusted.

Two further $90,000 awards cover the agent-specific risks: a malicious web page that, when Muse was asked to summarise it, mimicked status messages from the agent's browsing tool and got it to execute actions that should have been caught — Meta says its classifiers now catch the attack and 100 variants — and a formatting trick that slipped one-time passcodes past Muse's email redaction filters. A $50,000 award covered an Android flaw that let a malicious app load an attacker's page into a privileged web view, and $1,000 a Mac flaw that could hijack auto-update. For an agent that acts across a user's email, calendar and messaging accounts, the bounty table is a map of what can go wrong.

ANNOUNCED

Meta acquires Conductor, the multi-agent coding orchestrator — its cloud service shuts down on November 20

Meta has acquired Conductor, the startup whose Mac-native tool lets developers run several AI coding agents at once in parallel workspaces. Axios reported the deal first on November 4; founder and CEO Charlie Holtz confirmed it on Conductor's blog the same day, and Meta confirmed the acquisition to TechCrunch. Financial terms were not disclosed.

Conductor, founded in 2024 and launched in July 2025 after a $22 million Series A, builds no models of its own — it orchestrates other labs' agents, including Claude Code, OpenAI Codex, Cursor and OpenCode, each in an isolated workspace. The team is joining Meta Superintelligence Labs, working with chief AI officer Alexandr Wang and product head Nat Friedman, with something new promised for early 2027. Meta launched its own coding agent, Muse Code, in August and has reported more than 10 million users for its Muse agents.

For current users the clock is already running. Conductor has stopped new sign-ups and downloads; Conductor Cloud goes read-only on November 13 and shuts down permanently on November 20, after which cloud user data will be deleted. The local app keeps receiving new models and critical bug fixes until early 2027.