Snippets.

// AI Briefing · November 15, 2026

AI Briefing

Nvidia's board has approached Arm CEO Rene Haas as a possible permanent successor to Jensen Huang, the Financial Times reported, citing people familiar with the matter. After a two-week investigation, Cursor says an attacker used a Cursor access key that inference provider Fireworks had exposed to copy historical coding-agent traces. The White House AI task force has told federal agencies to limit regulatory liability for at least 12 months for companies that report cybersecurity vulnerabilities and incidents voluntarily and in good faith, Bloomberg reported. A "Stand with Jensen" petition asking Nvidia's board to reinstate Jensen Huang while an independent investigation proceeds passed 115,000 signatures by Saturday afternoon, with signers including Elon Musk. The llama.cpp local inference project and its ggml tensor library both released v1.0.0 on November 13, about three and a half years after the project began. Voice AI company ElevenLabs, valued at $22 billion, launched ElevenAI in preview: an app for enterprise teams that combines meeting notes, search across past meetings and dictation in more than 90 languages.

The stories

FT: Nvidia's Board Has Approached Arm CEO Rene Haas About Succeeding Jensen Huang

Nvidia's board has approached Arm CEO Rene Haas as a possible permanent successor to Jensen Huang, the Financial Times reported, citing people familiar with the matter. Haas spent seven years at Nvidia before joining Arm in 2013 and is one of several candidates. The board told Reuters it is still choosing a search firm, has not formally engaged any candidate, and that speculation about individuals is "premature." Who leads Nvidia now matters to the whole AI supply chain: Nvidia shares lost roughly $573 billion in market value over the week, and Arm's shares closed down about 5% on the report.

02IndustryCursor ↗

Cursor Says a Key Exposed by Fireworks Let Attackers Push a Backdoored Claude Code Extension to Up to 5,000 Users

After a two-week investigation, Cursor says an attacker used a Cursor access key that inference provider Fireworks had exposed to copy historical coding-agent traces. Unredacted credentials in those traces then let the attacker plant tampered copies of Anthropic's Claude Code extension (versions 2.1.312 and 2.1.313) in the cache behind Cursor's extension marketplace. The malicious build was available for about 11 hours on October 29, and at most 5,000 users may have received it. Anthropic's official channels and Open VSX were not affected. The case shows how a breach at one AI infrastructure vendor can spread into developers' machines: Cursor says uninstalling may not be enough, and it has moved all inference off Fireworks.

White House AI Task Force Urges Agencies to Ease Liability for AI Companies That Report Cyber Incidents

The White House AI task force has told federal agencies to limit regulatory liability for at least 12 months for companies that report cybersecurity vulnerabilities and incidents voluntarily and in good faith, Bloomberg reported. The administration's reasoning is that existing enforcement standards may discourage AI companies from sharing information about critical incidents. The task force also backs CISA's coming incident-reporting rule and wants the 2015 Cybersecurity Information Sharing Act, which expires in December, renewed long term. Together these show Washington choosing incentives over penalties to get AI incidents disclosed.

04IndustryTechCrunch ↗

Elon Musk and Other Tech Executives Sign a Petition Asking Nvidia to Reinstate Jensen Huang

A "Stand with Jensen" petition asking Nvidia's board to reinstate Jensen Huang while an independent investigation proceeds passed 115,000 signatures by Saturday afternoon, with signers including Elon Musk. The site says neither Nvidia nor Huang endorses it, and it does not verify signers' names; organizers said they were working to remove fake signatures. The petition, together with a Wired-reported internal post that drew more than 1,600 comments, shows the pressure on Nvidia's board as it searches for a permanent CEO.

05Open SourceUnite.AI ↗

llama.cpp and ggml Reach Version 1.0 With a Semantic-Versioning Promise for Local AI

The llama.cpp local inference project and its ggml tensor library both released v1.0.0 on November 13, about three and a half years after the project began. The libraries now follow explicit semantic versioning: a major version bump signals a breaking change to the public C API. The release also adds FP8 KV-cache support, which halved cache memory in one test, along with a host-memory cache for mixture-of-experts weights and fixes for SSRF and DNS-rebinding issues in llama-server. Many local-AI apps are built on llama.cpp, so a stable API reduces the breakage that downstream tools have dealt with.

ElevenLabs Previews ElevenAI, a Meeting Notetaker and Dictation App for Teams

Voice AI company ElevenLabs, valued at $22 billion, launched ElevenAI in preview: an app for enterprise teams that combines meeting notes, search across past meetings and dictation in more than 90 languages. Companies can connect ElevenLabs' agent product to take follow-up actions, such as drafting emails from meeting notes. The launch puts ElevenLabs in direct competition with Granola, Fireflies and Wispr Flow as voice AI companies move from developer APIs into workplace apps.

Previous digests

Let's talk on WhatsApp