Thursday, 29 Oct 2026
Subscribe to AIWatcher
AIWatcher
  • Home
  • News

    OpenAI’s Berlin pitch was about owning the runtime, not the model

    By
    AIWadmin

    A one-researcher model lab ships a 10M-token coding engine

    By
    AIWadmin

    Liquid AI branches six small models off one backbone

    By
    AIWadmin

    Google gives language models a sleep cycle to remember new facts

    By
    AIWadmin

    Harvey clears the federal security bar for legal AI

    By
    AIWadmin

    Musk predicts a cull of 200 neoclouds as debt markets close

    By
    AIWadmin
  • Articles

    Google hands Oracle the keys to sell its TPUs

    By
    AIWadmin

    Airbnb drags recommendation data from two days old to twenty seconds

    By
    AIWadmin

    Android Studio opens its doors to Claude, Codex and 40 other agents

    By
    AIWadmin

    A Delaware plan would let software run companies with nobody in charge

    By
    AIWadmin

    OpenAI halves its agent pricing with a cheaper vision model

    By
    AIWadmin

    Claude for Legal now checks its own citations first

    By
    AIWadmin
  • Spotlight

    Circuit Mind puts a swarm of agents on schematic review

    By
    AIWadmin

    Upwind turns down a takeover and raises more instead

    By
    AIWadmin

    DeepMind keeps Gemini 4 Carbon behind a cyber defenders gate

    By
    AIWadmin

    Nvidia folds three speech jobs into a single open model

    By
    AIWadmin

    An inference cloud bets on rival silicon while the bears circle

    By
    AIWadmin

    Publishing staff threw a party for an AI assistant that died

    By
    AIWadmin
  • About
    • Mission
    • Services
    • Contact
  • Newsletter
  • Shop
    • All Items
    • By Category
      • Hats
      • T-Shirts
    • Cart
  • 🔥
  • Alignment
  • Classification
  • Distillation
  • Explainability
  • Hallucination
  • Legal/Compliance
  • Medical
  • NLM
  • Mobility
  • Research
  • Robotics
  • Safety
  • Startups
  • Prompt
  • Python
  • RAG
  • RLHF
  • Token
  • Vision
Font ResizerAa
AIWatcherAIWatcher
  • Home
  • News
  • Articles
  • Spotlight
  • About
  • Newsletter
  • Shop
Search
  • Home
  • News
  • Articles
  • Spotlight
  • About
    • Mission
    • Services
    • Contact
  • Newsletter
  • Shop
    • All Items
    • By Category
    • Cart
Have an existing account? Sign In
Follow US
© 2022 Foxiz News Network. Ruby Design Company. All Rights Reserved.
Articles

AI labs turn to dead startups for training data

A new industry is brokering the code, chats, and documents of shuttered companies to labs that have exhausted public datasets.

AIWadmin
Last updated: October 15, 2026 2:37 am
AIWadmin
ByAIWadmin
Global AI news & information.
Follow:
Share
SHARE

When a startup folds, the leftovers it once intended to erase can turn out to be worth something after all. Brokers now move that material, code and Slack logs alike, into the hands of AI labs that need lifelike training data.

Take Keelan Miskell, who wound down his crypto company this year. Two and a half years of work he assumed was forgettable, he found, meant something to somebody. The code base alone brought roughly $25,000, with more to come from workspace data in Google, Notion, Figma and Slack, likely near $15,000.

SimpleClosure brokered that deal, having added data sales in April. Over 150 deals have cleared through its Asset Hub, which projects handing startups more than $10 million by the close of 2026. Labs began hunting for this material last year once public and synthetic corpora ran dry; a real archive holds roughness a manufactured one lacks, from judgment calls to disagreements to dates tied to events that truly occurred.

Sunset, a rival broker, prices a dying client’s holdings by headcount, operating years, data volume, and how its team used its tools. Founder and CEO Brendan Mahony reaches for a mill analogy, calling the operational exhaust sawdust most people assumed was worthless when a great deal of value hides in it. Valuations stay loose because no one has published a comparison of real-world training against the substitutes, so one buyer’s quote can differ wildly from the next.

Money moves regardless. Google beat data vendor Mercor in August to buy Spirit Airlines’ records out of bankruptcy for $10 million: 100 million emails, 500 million Teams messages, 176,000 employee files. Spirit’s flight attendants’ union fought the sale over privacy, and one attorney advising AI startups brands such data “legally radioactive.”

TAGGED:acquisitionsAI industryAI Training Datadata privacyStartups
SOURCES:MIT Technology Review
Share This Article
Email Copy Link Print
ByAIWadmin
Follow:
Global AI news & information.
Previous Article OpenAI and Anthropic back binding British AI rules
Next Article Stripped guardrails let open models answer terror prompts

You Might Also Like

News

Meta rewrites its business terms to feed AI on linked sites

By
AIWadmin
Articles

DeepMind proof system cracks open problems and clears peer review

By
AIWadmin
Articles

NVIDIA open-sources OSMO to run robot pipelines from one file

By
AIWadmin
Articles

IBM’s llm-d reaches 1.0 with controls enterprises kept asking for

By
AIWadmin
AIWatcher
Facebook Twitter Youtube Linkedin Rss

Global AI News and Information
AIWatcher is your definitive source for AI updates worldwide, from Silicon Valley to Shanghai.
Our industry coverage keeps you in the loop with the latest news and trends shaping the future of AI.

Quick Links
  • News
  • Articles
  • Spotlight
  • Events
About Us
  • Mission
  • Services
  • Contact
  • Privacy Policy
  • Legal
© 2026 AIWatcher. All Rights Reserved.