@Secludy
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States App Store
Account-level information from X, not a live location or the device used for a specific post.
de-ID for RL/data companies selling to AI labs - https://xitter.cf/t.co/QculTGI5iB
Joined September 2024
- Tweets14
- Following29
- Followers101
- Likes15
Secludy AI retweeted
if you sell AI training data, see how you compare: we found 159 companies selling data, RL envs and evals, every fact sourced
plus 38 programs that pay companies for their data: whosellsdata.com
8 of the 14 that pay companies directly publish a dollar range. most of the rest set terms per deal. unknowns are marked and nothing is ranked
the other page: what each program wants, who qualifies and what it offers: whosellsdata.com/buying
missing something, or something wrong? fix it on the site, a person reviews every change
we built this at @Secludy. if you buy, broker or license company data and need the names out before it reaches a lab, dm me
Secludy AI retweeted
100% agree on 3rd party de-id. we provide this at @Secludy. raw data going to the independent de-id provider before going to the data broker builds trust
This guy did $1,000,000 in 30 days as an AI data broker.
His model: frontier labs have used all the data on the internet and need private company data to keep improving.
His firm finds businesses with the right data then de-identifies it, packages it, and sells it to the labs.
@RyanLockAI runs one of the hottest startups in this space (Polyshares) and walked me through his entire business model on the pod.
Here's what I learned:
1. Making an intro to a data buyer is only step one. Technical due diligence on how the business actually operates is where the value gets built.
2. Labs buy the full workflow. Client email to internal Slack debate to resolution, including the fixes the team rejected.
3. He targets companies with 50+ employees, 3+ years on the same software, US-based, English. Below 50 employees = fixed legal and diligence costs kill the deal.
4. Remote-first white collar wins. If the work happens on the phone, the lab just sees "ticket created, ticket closed."
5. Depth of history beats raw volume. 20 years in business means nothing if they switched software 3 years ago.
6. Never let sellers de-identify their own data. Names become "employee 632" so relationships survive. DIY versions wreck the value.
7. Use a third party to de-identify. They hold the raw files for about a week, then delete them. That kills the biggest objection.
8. Never sign a perpetual license. 24 months of exclusivity is the typical buyer ask.
9. Time the exclusivity to the exit. Selling in 2 years? Match the term so the new owner inherits a dataset they can license again.
10. Demand rotates. A few weeks ago labs wanted heavy construction. Now it's video game companies.
His 2 key takeaways:
1. Act like a sell-side broker, not a lead seller. Organize and present the owner's data like an M&A deal and you can earn an advisory fee on top of the finder's fee.
2. The buyer pool could explode. If enterprises start training their own models, buyers go from maybe 10 labs to hundreds of thousands.
Ryan is doing this at a level most of the people hyping it on X are not, and we had a blast going deep on it. Go follow @RyanLockAI
(Also available on the Build With AI podcast wherever you get your pods)
Secludy AI retweeted
We're proud to be launching @Secludy today with $4M in seed funding.
@ImpressionVC led the round. @LAUNCH and The Syndicate, a venture firm and angel investing group led by @Jason Calacanis, also joined, along with Wedbush Ventures, @PrecursorVC, @HustleFundVC, @scriptcapital, Mana Ventures, Chispa VC, and an amazing group of angel investors.
Banks and fintechs are sitting on incredibly valuable data but can't use it to train AI models. Privacy laws, contracts, and cross-border rules keep it locked down.
So my co-founder @Dr_Mingze_He and I got to work.
Secludy unlocks your most sensitive data while keeping the utility intact. You can train custom models and do vendor evaluations without ever putting your sensitive data at risk or giving up AI model performance.
Moving on to other regulated industries next...
Also, special thank you to @sososazesh who wrote our first check and to all of our other existing investors.
Read more in the comments.
Secludy AI retweeted
The solutions already exist, they're just unevenly distributed (/think tanks are not aware of what startups are building to solve these problems)
See: @Secludy
1/7
High-quality data is a major bottleneck to AI progress.
But while recent LLMs were trained on ~hundreds of TBs of data, the world has digitized 180 zettabytes of it. A billion times more.
The problem is access.
In a new essay for The Launch Sequence, @iamtrask and @lace31692 lay out a possible solution…
👀
LLMs leak up to 27.5% of sensitive training data PII (Personally Identifiable Information like emails, SSNs, VINs, Bitcoin wallets).
@Secludy makes it easy to generate privacy-guaranteed synthetic data that is a near replica of the original unstructured dataset but better.
How? They utilize privacy-protected LLMs by adding carefully controlled noise to the model weights before generating synthetic data for AI model fine-tuning/evaluations. They just released a technical report that demonstrates their approach.
🧵1/n
LLMs are known to memorize and expose sensitive information, even when trained on masked unstructured datasets they can still retain and regurgitate personal data which is a major privacy risk.
We just launched our PII Leakage Testing Tool on AWS Marketplace!
We put data masking to the test. A thread 🧵👇
4/ The European Data Protection Board (EDPB) now wants companies to run these audits as proof of resistance to re-identification (EPDB Opinion 28/2024).
Secludy AI retweeted
Training AI models on sensitive medical data poses big risks, as this TechCrunch article highlights. The same goes for unstructured text. Privacy-preserving techniques like training on differentially private synthetic data can ensure utility while protecting privacy 🛡️ techcrunch.com/2024/11/19/ps…
