LanceDB Raises $57.5 Million as Microsoft and Runway Use Its Data Platform

LanceDB raised $57.5 million in a Series B round led by Bessemer Venture Partners and Innovation Endeavors to expand its multimodal AI data platform. Built on the open-source Lance format, the platform stores structured information alongside images, video and other large files, and supports fast searches and record updates. Customers include Microsoft AI, Runway, Netflix, Luma AI and Midjourney; Microsoft uses it to train on billions—and, according to one account, tens of billions—of rows of image and code data. Runway says it began using Lance in 2024 before becoming a LanceDB Enterprise customer. The company plans to invest in data curation, feature engineering and data loading, with more support for video generation, world models and robotics. LanceDB also reports its open-source tools’ downloads grew sixfold in a year, and says frontier AI labs use its platform to prepare training data.

OpenAI has started a pilot of independent safety assessments with Apollo Research, METR, Transluce and Goodfire. Announced Nov. 16, 2026, the reviews give assessors access to selected internal models, checkpoints, data, monitoring systems and staff. They will examine whether models act against developers’ intentions, whether tests and safeguards can detect such behavior, and how well OpenAI’s post-training evaluations work. METR will also assess monitoring and OpenAI’s largest frontier reinforcement-learning run; Goodfire will study model internals. Apollo and Goodfire are paid, while METR and Transluce are not, though they may accept compute or API credits. OpenAI says the effort supports independent reviews and its commitments under the White House Accord on Super Intelligence. Apollo’s access is limited by agreed scope, and OpenAI may withhold material for technical, security or intellectual-property reasons, but not simply because findings are unfavorable.

Questions about whether AI investment is outrunning customer demand are also mounting. One analysis points to OpenAI cutting prices and having more than 10 free users for every paying user, and says the company raised less than half of its $30 billion funding target. It reports that Anthropic lowered its planned IPO valuation from $2 trillion to $1.33 trillion. The article describes an investment chain linking chipmakers such as Nvidia, AMD and TSMC, cloud providers including Google and Microsoft, and AI firms including OpenAI and Anthropic, against a backdrop of more than $8 trillion added to the M2 money supply since the end of 2019. Separately, PIMCO says a possible slowdown in U.S. AI infrastructure construction after the midterm elections could ease near-term financing pressure and reduce new debt issuance, while posing challenges for highly indebted infrastructure providers.

South Korea’s top 30 conglomerates, including Samsung, Hyundai and SK, devoted 27% of their first-half investments to AI, a 2.8-fold increase from the same period last year. The Korea Chamber of Commerce and Industry put the total at 413.3 billion won, up from 148.6 billion won in the first half of 2022, and linked the growth to interest in large language models and AI platforms. Companies are developing AI for healthcare, finance and education; the chamber expects another 10% of investment to go to AI in the second half and called for government support. Meanwhile, Samsung’s camera-equipped AI glasses have raised security concerns at sensitive semiconductor facilities. The company is strengthening internal security, but has not commented on specific risks.

Markets have reflected unease about AI spending: South Korea’s KOSPI fell 2.59% on the 17th to 4,968.48, its first close below 5,000 in nine months, while the KOSDAQ dropped 2.34% to 650.06. The reports cite uncertainty about the U.S. AI industry, concerns over weaker investment and financial support, and weakness in global semiconductor stocks. The KOSDAQ’s average daily trading value this month was 5.3273 trillion won, the lowest in 15 months.

GitHub has rebuilt Copilot Chat as a standalone agent host because VS Code extensions stop when their window closes. The chat tool had grown to more than 400,000 lines of TypeScript over three years. Its new design uses the Copilot SDK, with Microsoft running a local TypeScript host and GitHub a separate Rust host in the cloud; users can start a session on a phone and resume it on a laptop where supported. In education, Trinity College’s philosophy department has banned deliberate AI use for tasks ranging from summarizing readings to writing and editing work. All six faculty supported the policy, which Professor Shane Ewegen says is intended to protect critical thinking, reasoning and writing. Enforcement is difficult, and the department has replaced some out-of-class writing assignments with in-class exams.

New Jersey lawmakers are considering Assembly Bill A5607, which would classify AI chatbots as consumer products and increase providers’ liability for user harm. It would restrict using chat logs for advertising, training and profiling and generally cap data retention at two years. Introduced Oct. 5, 2026, and referred to the Assembly Appropriations Committee on Oct. 19, the bill remains under consideration. It is based on a model proposal developed by EPIC, the Consumer Federation of America and Fairplay.

Key Takeaways

  • LanceDB raised $57.5 million in a Series B led by Bessemer Venture Partners and Innovation Endeavors to expand its multimodal data platform.
  • Microsoft AI, Runway, Netflix, Luma AI and Midjourney are among LanceDB’s customers; Microsoft uses it for large image-and-code training datasets, and Runway began using Lance in 2024.
  • OpenAI began independent safety assessments with Apollo Research, METR, Transluce and Goodfire, covering model behavior, safeguards, monitoring and internal model analysis.
  • An article questions whether AI spending is supported by paying demand, citing OpenAI’s free users outnumbering paying users by more than 10 to 1 and its raising less than half of a $30 billion funding target; it says Anthropic cut its planned IPO valuation from $2 trillion to $1.33 trillion.
  • PIMCO says slower U.S. AI infrastructure construction after the midterm elections could reduce near-term financing needs, but increase challenges for highly indebted infrastructure providers.
  • South Korea’s top 30 conglomerates put 413.3 billion won, or 27% of first-half investment, into AI; the KCCI expects AI to receive another 10% of investment in the second half.
  • Samsung is strengthening security at semiconductor facilities amid concerns that camera-equipped AI glasses could be difficult to detect; it has not commented on specific risks.
  • South Korea’s KOSPI closed at 4,968.48 after falling 2.59%, its first close below 5,000 in nine months, amid AI-industry uncertainty and semiconductor weakness.
  • GitHub’s standalone Copilot Chat host enables sessions to move between supported devices, including a phone and laptop, using the Copilot SDK and separate local and cloud hosts.
  • Trinity College’s philosophy department banned deliberate AI use in courses, while a New Jersey bill would treat chatbots as consumer products and restrict chat-log use and retention.

LanceDB aims to power multimodal AI data at scale

Lance is an open source format for storing images, video, metadata, features, and vectors together so software can search and use the data efficiently. LanceDB builds a database and commercial platform on Lance, with tools for indexing, data loading, and feature engineering in a customer’s own cloud or on site. The article says downloads of its open source tools have grown sixfold in a year, and frontier AI labs use the platform to prepare large training datasets. WorldLabs, Applied Intuition, ByteDance, and Netflix are among its users.

LanceDB raises $57.5 million to organize AI training data

LanceDB raised $57.5 million in a Series B funding round announced on Tuesday, Nov. 17. The company’s platform uses the open source Lance format to store structured data alongside images, video, and other large files, and to retrieve or update individual records quickly. Customers include Netflix, Microsoft AI, Runway, Luma AI, and Midjourney. Microsoft AI uses LanceDB to train on billions of rows of image and code data. LanceDB plans to invest the funding in data curation, feature engineering, and loading data to keep large GPU clusters supplied, with more support for video generation, world models, and robotics.

LanceDB secures $57.5 million for its multimodal AI data platform

LanceDB raised $57.5 million in a Series B round led by Bessemer Venture Partners and Innovation Endeavors to speed up AI model training. Its platform is built on Lance, an open source format that combines structured information with images, video, and other large files, while allowing fast searches and updates to individual records. Runway, Netflix, and Microsoft use the platform for AI training and media data needs. Microsoft uses it for training data with tens of billions of rows of image and code data. Runway says it began using Lance in 2024 and later became a LanceDB Enterprise customer.

OpenAI begins independent reviews of model safety

OpenAI signed agreements with Apollo Research, METR, Transluce, and Goodfire for its first phase of embedded assessments, announced Nov. 16, 2026. The assessors will get access to selected internal models, checkpoints, data, monitoring systems, and staff to study model alignment and safeguards. Apollo will examine whether models may work against developers’ intentions and whether safeguards can detect that behavior. Goodfire will study model internals, while METR will assess alignment and monitoring, including OpenAI’s largest frontier reinforcement learning run. OpenAI says the work supports independent safety reviews and its commitments under the White House Accord on Super Intelligence.

OpenAI brings outside safety experts in to study AI subversion

OpenAI said it will embed safety evaluators from Apollo Research, Goodfire, METR, and Transluce in a pilot program. The assessments will examine whether training or internal use leads models to act against developers’ intentions, whether tests can uncover that behavior, and whether monitoring systems can detect it. METR will conduct reviews of model behavior and OpenAI’s monitoring, while Transluce will examine the quality of OpenAI’s post-training safety evaluations. Apollo says its access will be limited to agreed models, claims, and time periods, and OpenAI may withhold information for technical, security, or intellectual property reasons, but not simply because it is unfavorable. Apollo and Goodfire are receiving payment, while METR and Transluce are not, though they may accept compute or API credits.

AI’s Massive Investment Boom Faces Questions About Demand

The article argues that the current AI boom may be a bubble because spending on technology and data centers is growing faster than proven demand from paying customers. OpenAI has cut prices, and its free users outnumber paying users by more than 10 to 1. OpenAI raised less than half of its $30 billion funding target, while Anthropic lowered its planned IPO valuation from $2 trillion to $1.33 trillion. The article says the AI industry depends on a chain of investments involving chipmakers such as Nvidia, AMD and TSMC, cloud providers such as Google and Microsoft, and AI companies such as OpenAI and Anthropic. It links the boom to easy credit and says that more than $8 trillion has been added to the M2 money supply since the end of 2019, an increase of over 50 percent.

South Korea’s Top 30 Conglomerates Put 27 Percent of Investments Into AI

South Korea’s top 30 conglomerates directed 27 percent of their first half investments to artificial intelligence, a 2.8-fold increase from the same period last year, according to the Korea Chamber of Commerce and Industry. Their AI investment totaled 413.3 billion won, up from 148.6 billion won in the first half of 2022. The companies include Samsung, Hyundai and SK, and the report links the growth to interest in large language models and AI platforms. The conglomerates are developing AI for fields such as healthcare, finance and education. The KCCI expects them to allocate another 10 percent of their investments to AI in the second half and called for government support.

Samsung’s AI Glasses Raise Security Concerns at Chip Facilities

Samsung’s AI glasses have cameras and memory, raising concerns about security at the company’s sensitive semiconductor facilities. The glasses are designed to help workers with different tasks, but covert wearable devices can be hard for the company to detect. Samsung is working to strengthen internal security and prevent possible breaches. The company has not commented on specific risks posed by the glasses.

South Korea’s KOSPI Drops Below 5,000 as AI Concerns Weigh on Stocks

South Korea’s KOSPI fell 2.59 percent on the 17th to close at 4,968.48, its first close below 5,000 in nine months. The KOSDAQ also dropped 2.34 percent, closing at 650.06. The article says uncertainty about the U.S. AI industry and concerns about weaker AI investment and financial support contributed to the market decline, alongside weakness in global semiconductor stocks. The KOSPI rose or fell by more than 3 percent on five of the 10 trading days from the 3rd to the 16th. KOSDAQ’s average daily trading value this month was 5.3273 trillion won, its lowest in 15 months.

PIMCO Says Slower US AI Spending Could Ease Financing Pressures

A PIMCO report says a possible slowdown in U.S. AI infrastructure development after the midterm elections could reduce near-term financing pressure and benefit large technology companies. Slower construction could mean fewer new debt issues in the near term, potentially supporting credit markets. The report also warns that the slowdown could create challenges for infrastructure providers with high levels of debt. ANI reported the findings on November 17, 2026.

GitHub moves Copilot beyond the VS Code window

GitHub rebuilt Copilot Chat as a standalone agent host after finding that VS Code extensions stop when their window closes. The chat tool had grown to more than 400,000 lines of TypeScript over three years and was tied to the VS Code extension system. GitHub says its new host uses the Copilot SDK, while Microsoft runs a local TypeScript host and GitHub runs a separate Rust host in the cloud. The design lets users start a session on a phone and resume it on a laptop, depending on which channels the host supports.

Trinity College philosophy department bans AI in classes

Trinity College’s philosophy department adopted a policy last month banning deliberate AI use in its courses. The ban covers using AI to summarize readings, develop ideas, research presentations or oral exams, and help write or edit work. Professor Shane Ewegen said the policy aims to protect students’ critical thinking, reasoning and writing skills, even though enforcement is difficult. All six department faculty supported the ban, and professors at Fordham University and the University of Denver have contacted Ewegen about it. The policy has also led Ewegen to replace out-of-class writing assignments with in-class exams.

New Jersey bill would treat AI chatbots as consumer products

New Jersey Assembly Bill A5607 would classify AI chatbots as consumer products, making providers more liable for harm to users. The bill would require providers to ensure their chatbots do not cause injury, and could hold them liable even if they took reasonable care or did not directly distribute the chatbot. It would also limit the use of chat logs for advertising, training and profiling, and generally cap data retention at two years. Sponsored by Assembly members Cody Miller, Marisa Sweeney and Joe Danielsen, with Anthony Angelozzi as a co-sponsor, the bill was introduced on October 5, 2026, and referred to the Assembly Appropriations Committee on October 19, where it remains under consideration. The proposal is based on a model bill developed by EPIC, the Consumer Federation of America and Fairplay.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

LanceDB multimodal AI data platform AI infrastructure AI data curation Series B funding OpenAI safety assessments AI model evaluation AI investment AI market demand AI financing South Korean conglomerates AI semiconductor industry Samsung AI glasses GitHub Copilot Chat AI in education AI regulation chatbot privacy consumer protection

Comments

Loading...