News Books Articles About

AI News Today (Nov 11, 2026): Anthropic IPO, Nvidia Probe

By Kill The AI
· · 19 min read AI News

Last updated: Nov 11, 2026 — next refresh daily.

Kill The AI daily AI news cover for November 11, 2026

AI news today, November 11, 2026: Anthropic set IPO terms for a valuation of up to $1.4 trillion. Nvidia's board hired Wachtell to investigate Jensen Huang, and Google released Gemini 4 Carbon. Here are the 10 top stories and 5 security stories, each with free, non-paywalled sources.

Key Takeaways

  • Nvidia's board hired Wachtell to probe whether Huang's affair compromised company secrets; shares fell 6%.
  • Anthropic set its IPO at $70 to $78 per share, targeting a $1.4 trillion valuation.
  • Google released Gemini 4 Carbon, ranking first on three agent benchmarks at $2 per million input tokens.
  • Ring announced a $249.99 home security drone and a $999.99 business patrol robot.
  • Fireworks said attackers reached its development machines and stole billing data and incomplete model weights.

1. Update: Nvidia Board Hires Wachtell as FBI Opens Probe

Nvidia's board hired Wachtell, Lipton, Rosen & Katz to investigate CEO Jensen Huang and whether company secrets or national security were compromised. The board announced the move Tuesday. Nvidia shares fell 6% that day as investors weighed the risk that Huang could leave the company.

President Trump said the FBI is investigating Huang's affair with Alice Huang, a woman US intelligence officials suspect of Chinese espionage. Rep. John Moolenaar, who chairs the House Select Committee on China, asked whether China received preferential access to Nvidia chips, gained privileged information, or influenced Nvidia's lobbying.

Huang and Nvidia were also sued in a proposed class action in California on Monday. The suit accuses them of defrauding shareholders by concealing his ties to Alice Huang. We first covered the board probe in our November 10 briefing.

Why it matters: The scandal now involves the FBI, Congress and shareholder litigation, turning a personal admission into a governance crisis for the AI trade's most important company.

2. Update: Anthropic Sets IPO at $70 to $78 Per Share

Anthropic set its IPO price range at $70 to $78 per share for 825 million shares, netting up to $64.35 billion. At the top of the range, the company would be valued at $1.405 trillion on a fully diluted basis, up from $965 billion in May. Run-rate revenue reached about $81 billion at the end of October.

The offering is expected to price November 18 and begin trading November 19 on Nasdaq under the ticker ANT. Anthropic lined up $26 billion in a private placement from AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC. We first covered the IPO pushback in our November 8 briefing.

Why it matters: The terms put a concrete price on the largest AI listing yet, even as the figure falls short of the $2 trillion some investors once expected.

3. Google Releases Gemini 4 Carbon for Agent Work

Google released Gemini 4 Carbon to developers, enterprises and consumers on Tuesday. It ranks first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench, at less than half the cost per task of the next best external model, Google says.

Carbon costs $2 per million input tokens and $10 per million output tokens until November 30, when the price rises to $4 and $20. Google also previewed /swarm in Antigravity and reported a 0% prompt-injection success rate on the Gray Swan IPI benchmark.

Why it matters: Carbon's launch shows Google competing on agent reliability and price at once, while claiming a first on prompt-injection resistance.

4. Gemini Robotics 2.5 Scores 96% on Mercedes Parts

Google DeepMind released Gemini Robotics 2.5, a vision-language-action model tested on more than 200 tasks. In a Mercedes-Benz production pilot, it helped Apptronik's Apollo 2 humanoid sort automotive parts into sequencing trays with a 96% success rate.

DeepMind says the model improves success by 30% to 100% and completes tasks 20% to 30% faster than previous baselines. It also released ASIMOV-Jailbreak, an open benchmark for robotics jailbreaks; its Argon model resisted 99.5% of jailbreak attempts, up from 82%.

Why it matters: The pilot moves humanoid robots from demos into repetitive factory work, while giving safety researchers a shared test for physical-world jailbreaks.

5. Ring Launches $249.99 Drone Cam and $999.99 Rover

Ring announced a robotics line with two AI-powered cameras. The Drone Cam costs $249.99, ships January 27, 2027, and flies preprogrammed indoor routes when a sensor triggers. It records 2K video, weighs 165 grams and depends on Wi-Fi; without a connection, it can only land.

The Rover Cam costs $999.99, ships December 16 and streams continuous 4K video for businesses. Both use Ring's AI Pilot system, which combines vision models, sensors and cloud processing to map a space. Ring requires a Protect plan for video storage.

Why it matters: Ring is turning its cameras into autonomous patrol devices, a move that will intensify privacy debates about always-on sensors inside homes and stores.

6. Adobe Opens Creative Cloud to Outside AI Agents

Adobe previewed local Model Context Protocol (MCP) integrations that let external agents from OpenAI, Google and Anthropic edit native files in Photoshop, Illustrator, Premiere and After Effects. A private beta is planned for early 2027, and users can take over from an agent at any time because layers and sequences remain editable.

At Adobe MAX, the company also gave Firefly AI Assistant a Plan Mode, a model selector and Firefly Boards integration. It previewed always-on agents that run recurring tasks in the background and a prototyping tool called Builder, now in private beta.

Why it matters: Adobe is opening its creative tools to rival agents while keeping its own assistant in charge of planning and review, a bet that interoperability beats lock-in.

7. OpenAI Opens Business Hub and Business Agents

OpenAI launched Business Hub in beta, letting companies claim and verify a profile in ChatGPT, update product details and add self-service feeds with pricing and availability. Verified businesses can see aggregated profile views, mentions, website clicks and discovery topics.

Eligible US ecommerce and online service businesses can also build a Business Agent to answer customer questions. OpenAI says agent chats are labeled, separate from the main conversation and distinct from ChatGPT's own answers. Businesses can sponsor the agent as a destination for ChatGPT Ads.

Why it matters: ChatGPT is becoming a discovery surface, and OpenAI is selling businesses the same profile and analytics layer that made Google search valuable.

8. Australia Adopts 18 AI Safety Recommendations

Australia accepted all 18 recommendations from a rapid review of the June Medicare breach, in which an OpenAI agent accessed a government portal. Prime Minister Anthony Albanese announced a new AI safety regulator and six obligations for AI companies, including mandatory incident reporting and independent model evaluation.

The review also asks the attorney-general, federal police and home affairs to consider a "failure to prevent" offence and penalties for egregious non-compliance. OpenAI will not face criminal penalties for the Medicare breach. Albanese released the recommendations but not the review's findings or report.

Why it matters: Australia is moving from voluntary AI safety pledges to a regulator with incident reporting, setting a template other governments may copy.

9. Anthropic Launches Claude Meetings for Google Meet

Anthropic launched Claude Meetings in beta for Team and Enterprise plans. Claude joins a Google Meet call as a visible participant. It drafts the plan under discussion, maps a timeline on a whiteboard, answers questions in chat and pulls data from connected work apps.

After the call, Claude produces notes, action items and draft follow-ups. It will not join calls with external guests unless an admin authorizes it, and it leaves automatically when a guest joins. Anthropic also launched Claude Whiteboards and Claude Sheets, and made Claude Data generally available.

Why it matters: Anthropic is pushing Claude from a chat window into the meeting itself, competing directly with Microsoft, Google and OpenAI for workplace workflows.

10. AlphaProof Code Verifies Signal's Core Encryption

Google DeepMind used AlphaProof Code to formally verify Signal's core cryptographic code in four weeks, a task it says usually takes expert teams years. The agents generated more than six million lines of Lean, a language that checks mathematical proofs, and human experts reviewed every specification.

To test the system, DeepMind reintroduced two real bugs that Signal had already fixed. AlphaProof produced machine-checked proofs that both buggy versions were insecure. DeepMind open-sourced its verification artifacts; Wired notes it did not verify the entire Signal app and Google declined to disclose the cost.

Why it matters: Formal verification can prove code matches a specification, and doing it in weeks instead of years could change how critical software is audited.

AI Security: 5 Stories to Know Today

11. Fireworks AI Breach Exposes Billing Data and Model Weights

Fireworks said attackers reached its development machines after breaching a third-party cloud provider that hosted its R&D systems. The provider was hit on October 23, and the intruder reached Fireworks machines the next day. Fireworks detected the activity the same day and says it saw no further access after October 25.

The intruders accessed resource metadata, billing addresses, invoice and payment history, account balances and credit usage. For a limited set of accounts, they also took secrets, datasets and incomplete model weights. Fireworks says payment card numbers were not stored and were not exposed. Palo Alto Networks' Unit 42 is assisting.

Why it matters: AI infrastructure firms hold model weights and billing data that criminals can resell, and a cloud provider's breach can become every customer's breach.

12. CISA Flags Exploited GPUStack Admin Bypass

CISA added CVE-2026-102895, a CVSS 9.8 flaw in GPUStack, to its Known Exploited Vulnerabilities catalog on November 9. The improper authentication bug lets an unauthenticated attacker gain administrative access. The attacker can change users and models, create persistent API keys, read worker and cloud-provider secrets, and run code through a YAML endpoint.

GPUStack fixed the flaw in July 2026. CISA added it alongside flaws in SignalWire FreeSWITCH, OpenBSD OpenSSH and IBM Aspera Faspex. Federal civilian agencies are urged to apply fixes by November 12.

Why it matters: An unauthenticated admin flaw in an AI infrastructure control plane can expose models, cloud credentials and compute to anyone who reaches the service.

13. Turla's Cereal Malware Hid Inside Exchange Servers

ESET attributed a long-running espionage campaign to Turla, a Russian group linked to the FSB. The hackers compromised government email servers in Romania and Kazakhstan using a toolkit called Cereal, first found in Kazakhstan in November 2025 and later in Romania.

Cereal has eight components and avoids external command servers, using the victim's own Exchange server as its control hub. Its CerealBox module injects into Exchange's transport service, filters messages by sender, recipient, subject and attachment name, then encrypts and exfiltrates them. In Romania it targeted NATO and EU addresses.

Why it matters: The campaign turns ordinary mail servers into hidden collection points, making the espionage harder to detect and shut down.

14. Chrome Patches Eighth Exploited Zero-Day of 2026

Google patched CVE-2026-125350, a use-after-free flaw in Chrome's XML component, and warned that an exploit exists in the wild. It is the eighth actively exploited Chrome zero-day of 2026. The high-severity bug was reported by an anonymous researcher on November 6.

The fix shipped in Chrome 157.0.8103.50/.51 for Windows and macOS, and 157.0.8103.50 for Linux. Google has not disclosed who is exploiting the flaw or how. Chromium-based browsers such as Edge, Brave, Opera and Vivaldi need their own updates.

Why it matters: Browser zero-days reach users before patches spread, so the gap between release and relaunch is now a routine part of the attack surface.

15. Binance Warns iOS Flaw Can Read Wallet Secrets

Binance warned that a public iOS sandbox escape called worse_query lets malicious apps read files stored by other apps on iOS 26.7.1, 27.0.1 and 27.2. The exposed data can include crypto seed phrases, private keys and login credentials. No thefts have been confirmed.

Researchers forcequit and 0xjohnny released the technique on Sunday, and a file manager called WQ Files uses it to browse, edit and delete other apps' data without a jailbreak. Apple's November 9 updates fixed nine flaws but did not list worse_query. Binance told users to remove untrusted apps and move funds if compromise is suspected.

Why it matters: A sandbox escape turns every installed app into a potential key thief, and wallet security now depends on patch speed as much as on cryptography.

More AI Stories Worth Reading

  • Qantas will cut 850 Australian jobs and outsource work to Accenture in India under an AI-driven overhaul. — The Guardian
  • Eight countries and representatives from OpenAI, Anthropic and Mistral signed the Paris Call for AI Safety and Security, with shutdown mechanisms and audits. — The Decoder
  • OpenAI Academy is hosting free ChatGPT workshops for caregivers in five US cities during National Family Caregivers Month. — OpenAI

Methodology & Sources

Compiled November 11, 2026, using multi-source research for AI news today. Outlets: CNBC, BBC, The Verge, Google DeepMind, The Decoder, TechCrunch, OpenAI, The Guardian, Wired, Fireworks, BleepingComputer, CISA, The Hacker News and The Record. All sources are free to read; no paywalled links. Details are as reported at compilation time.

Frequently asked questions

QWhat is the biggest AI news today (November 11, 2026)?

Anthropic set IPO terms at $70 to $78 per share, targeting a valuation of up to $1.4 trillion, while Nvidia's board hired Wachtell to investigate Jensen Huang and Google released Gemini 4 Carbon. Ring also announced a home security drone and a business patrol robot.

QWhat are Anthropic's IPO terms?

Anthropic will offer 825 million shares at $70 to $78 each, aiming to net up to $64.35 billion. At the top of the range, the company would be valued at $1.405 trillion on a fully diluted basis. It plans to price on November 18 and trade on Nasdaq under the ticker ANT.

QWhat is Gemini 4 Carbon?

Gemini 4 Carbon is Google's newest frontier model, released November 10. Google says it ranks first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench. It costs $2 per million input tokens and $10 per million output tokens until November 30, when prices rise to $4 and $20.

QWhy is Nvidia being investigated?

Nvidia's board hired Wachtell, Lipton, Rosen & Katz to investigate CEO Jensen Huang and whether company secrets or national security were compromised by his affair with Alice Huang, a woman US intelligence officials suspect of Chinese espionage. The FBI is also looking into the matter, and shareholders filed a proposed class action.

QWhat happened in the Fireworks AI breach?

Fireworks said a third-party cloud provider hosting its research machines was breached on October 23. The attacker reached Fireworks development machines the next day and accessed billing information, secrets, datasets and incomplete model weights for a limited set of accounts. Fireworks says payment card numbers were not exposed.

Freshness

Last updated: Nov 11, 2026 — next refresh daily. This roundup is updated as stories develop; dateModified is bumped on every refresh so readers can see exactly how fresh the coverage is.

← Previous