All posts
#AI today#AI news#Cursor·November 16, 2026·AI-assisted

AI today: Cursor explains its breach, Claude starts joining Google Meet calls (16 Nov 2026)

Cursor says its October breach started with an access key exposed by its AI vendor Fireworks, Claude Meetings rolls out in beta for Google Meet, LangChain raises $300 million, Google opens its /teamwork agents to outside researchers, Apple starts a health AI study, and Delhi's Bharat Semi raises $24 million.

Arthat Team
4 min read
AI today: Cursor explains its breach, Claude starts joining Google Meet calls (16 Nov 2026)

Cursor has published what went wrong in its October security incident: an attacker got hold of an access key Cursor had issued to Fireworks, its AI training and inference vendor, and worked from there to a malicious copy of Anthropic's Claude Code extension in Cursor's marketplace. Cursor says it believes the incident is contained: the last attacker activity it saw was on 1 November, and it was blocked.

Cursor says its breach started with a key held by its AI vendor

On 31 October we covered the tampered Claude Code extension and asked how it got there. In its update of 13 November, Cursor says the key, which it says was "exposed by Fireworks", let the attacker copy a read-only bucket of coding-agent traces, mostly from 2025, from sessions not in Privacy Mode. A small number of traces held credentials that had not been redacted, and from those the attacker reached some of Cursor's SaaS tools and the cache behind its extension marketplace. Cursor says coding-agent traces were exposed for fewer than 0.05% of users, less than 0.01% of customers had API credentials exposed, at most 5,000 users may have received the malicious extension, and it has moved all inference off Fireworks as a precaution.

Fireworks' own update of 9 November does not name Cursor. It says a third-party cloud provider hosting its R&D machines had a security incident on 23 October, and the next day an unauthorised party used it to reach Fireworks development machines. A Fireworks spokesperson told Bloomberg the company is "actively working with our customers to manage the incident, including Notion, Cursor and others", Insurance Journal reports.

Why it matters: some of the agent traces Cursor shared with its vendor held credentials that had not been redacted. If you share AI logs with a vendor, check what is in them and strip credentials before they leave.

Claude Meetings joins Google Meet calls

Anthropic began rolling out Claude Meetings in beta on 12 November for Team and Enterprise plans, its release notes say. Claude joins a Google Meet call as a visible participant, takes notes, answers questions and builds docs, whiteboards and slides during the call. The help page says it saves a transcript, summary and screen descriptions, not audio, and asks you to make sure everyone on the call is okay with being transcribed.

Claude Meetings default settings on Team and Enterprise plans
Claude Help Center, checked 16 Nov 2026.

Why it matters: on Team plans it is on by default and allowed on calls with outside guests, so check the owner settings before your staff add it to client calls.

LangChain raises $300 million

LangChain raised a $300 million Series C at a $3 billion valuation, led by CapitalG, TechCrunch reports. The company says its LangSmith software for building and managing AI agents is not trying to lock users into any particular model, cloud or ecosystem, and claims 7,000 LangSmith customers.

Why it matters: if you are choosing a tool to build agents, check whether it lets you change the model or cloud underneath later.

Google opens its /teamwork agents to outside researchers

Google Research says its /teamwork feature in Antigravity runs several agents on one problem, gives each only the files and tools it needs, and checks candidate answers with "falsifiers" such as proof checkers and unmodified test suites. External scientists and developers can use it with Gemini 4 models today, Google says, with more features coming to outside users in the next few weeks. Google has also opened applications for 10,000 academic researchers to get six months of free access to its highest usage tier.

Why it matters: the part to copy is the checking step. If AI agents write code or analysis for you, run it through tests the agents cannot change before you use it.

Apple starts a study of how people use AI for health

Apple launched a study app with OpenEvidence for US users over 21 on iPhone 15 or later, running iOS 26 or later with Apple Intelligence on, aiming for up to 200,000 participants over more than 12 months, TechCrunch reports. Answers are grounded in sources including the New England Journal of Medicine, JAMA and Nature, and users rate them. Apple says data is anonymised and not used to train AI models, and OpenEvidence deletes questions and answers after 24 hours.

Why it matters: a clinic planning a patient chatbot can borrow Apple's setup: answers tied to medical sources, ratings from real users, and questions deleted after a short time.

Delhi's Bharat Semi raises $24 million for AI image sensors

New Delhi-based Bharat Semi raised a $24 million seed round led by Info Edge Group, South Park Commons and Indus Bridge Ventures, TechCrunch reports. It is building sensors that capture images, process them and run AI on one chip, which co-founder Vrinda Kapoor says could cut system costs by at least 70% compared with using a separate sensor, image processor and system-on-chip. The startup is pre-revenue and targets first commercial production in 2027.

Why it matters: the startup is aiming at smart homes, industrial automation, surveillance and aerospace. If you are planning camera-based automation, its first commercial production is targeted for 2027.

Worth watching: Microsoft Ignite opens in San Francisco on 17 November.

Arthat AI

Working on something similar at your company?

Arthat AI builds these automations for Indian businesses every week. If you want to talk through whether this fits your situation, we'll give you a straight answer — even if the answer is "don't build it."

Ready to make AI part of how your business works?

Book a 30-minute discovery call. We’ll listen, tell you where AI pays off first, and be honest if it doesn’t.

See what we’ve built