Your proprietary data is what AI labs are looking for.
Leading AI labs license proprietary data to train their models. Verilex prepares your data, clears the rights, and places it with the labs that need it, on terms your counsel can stand behind.
Direct access to AI labs
We present your data to the labs best matched to it and compare real offers side by side, so you license on the strongest terms.
PII handled before it's a liability
Detection, redaction, and verification are part of the process, not an afterthought left to the lab.
One accountable team
Rights review, preparation, negotiation, and delivery in one engagement, coordinated with your legal and security teams.
How an engagement works
From first conversation to signed license.
Data assessment
We review what you hold: volume, format, provenance, freshness, and exclusivity. We tell you plainly whether it has licensing value and which AI labs it fits.
Rights & risk review
Before anything is shown to an AI lab we confirm you own or control the rights to license it: customer contracts, terms of service, consent, third-party content, and regulatory limits.
PII scrubbing & preparation
Personal and confidential data is detected, redacted, or tokenized, then sampled and verified. Output is packaged in the formats AI developers actually ingest.
Place with AI labs
We present a de-identified sample and summary to the AI labs best matched to your data and put the offers side by side: price, exclusivity, term, and payment structure, not just headline numbers.
Negotiate & close
We help you negotiate license scope, exclusivity, field-of-use limits, audit and deletion rights, and indemnities, working alongside your counsel.
Deliver & monitor
Secure delivery, acceptance testing against the agreed spec, and follow-up to confirm the lab is staying within the license.
What data deals look like
Pricing is opaque and most contracts are confidential, but a handful of deals are public. These are benchmarks of the market, not quotes for your data.
Publicly reported deals
| Deal | Data | Reported value | Note |
|---|---|---|---|
| News Corp ↔ OpenAI | News and archive licensing | $250M+ over 5 years (~$50M/yr) | Largest disclosed content deal |
| Reddit ↔ Google | User-generated discussion data | ~$60M/yr (reported) | Part of $200M+ in contracts disclosed at IPO |
| Shutterstock ↔ Big Tech | Image, video and music library | $25M–$50M per deal | Per company CFO; smaller deals near $5M |
| Wiley | Academic publishing content | $44M AI revenue in 2024 | Company-confirmed; incl. a $23M one-time deal |
Typical per-unit rates (media content, 2024 reporting)
| Data type | Reported range | Note |
|---|---|---|
| Images / stills | $0.02 – $1 per image | Bulk archives sit at the low end; curated or specialized sets higher |
| Short video clips | $1 – $4 per clip | |
| Long-form video | ~$100 – $300 per hour | Adobe's call for footage worked out near $180/hr |
| Text | ~$1 per 1,000 words | Commodity text is far cheaper; expert or proprietary text far dearer |
What moves the price
- Uniqueness: data a lab can't get anywhere else
- Exclusivity and term: exclusive and long licenses cost more
- Clean rights: clear ownership and consent, with PII removed
- Quality and labeling: expert annotation, structure, freshness
- Field of use: training only vs. redistribution or resale
Read these numbers carefully
- Headline deals are with major publishers and platforms. Most companies will see far less.
- Many figures are press reports, not confirmed by the parties.
- Commodity data (public web text, generic photos) is worth little; scarce, rights-clean domain data is where value concentrates.
- Structures vary: flat fee, annual license, per-unit, or revenue share.
Sources: press and trade reporting including the Wall Street Journal, Reuters, and company statements, as compiled by LLM Pulse, Monda and The Decoder. Figures are from 2024–2025 reporting and may be out of date. Nothing here is an offer or a valuation of your data.
Data AI developers license
- ✓Customer support transcripts & chat logs
- ✓Domain-specific documents (legal, medical, financial, engineering)
- ✓Call-center audio & meeting recordings
- ✓Code repositories & technical documentation
- ✓Images, video & sensor / telemetry data
- ✓Proprietary databases & structured records
- ✓Archives: publishing, email, internal knowledge bases
- ✓Expert-annotated or labeled datasets
What we scrub
- ✓Names, addresses, emails, phone numbers, government IDs
- ✓Financial account and payment card numbers
- ✓Health information (HIPAA-regulated data and PHI)
- ✓Employee, customer, and vendor records
- ✓Credentials, API keys, and internal secrets
- ✓Free-text fields where identifiers hide in plain sight
Scrubbing is risk reduction, not a legal guarantee. Compliance obligations (GDPR, CCPA/CPRA, HIPAA, contractual limits) remain yours to confirm with counsel.
Where companies' data has value
Examples by industry of what is commonly licensed, who typically owns the decision, and what needs care before a deal.
| Industry | Data commonly licensed | Typical decision-makers | Watch for |
|---|---|---|---|
| Healthcare & life sciences | De-identified clinical notes, imaging, real-world evidence | Chief Data Officer, Head of Real-World Data, Privacy / Compliance | HIPAA, consent, de-identification standard |
| Financial services | Transaction patterns, research, alternative data, filings analysis | Chief Data Officer, Head of Data Products, Alternative Data lead | Customer-data terms, GLBA, model-risk rules |
| Media & publishing | Archives, articles, books, images, audio, video | Head of Licensing, Syndication, Chief Content Officer | Third-party and contributor rights |
| Software & SaaS | Documentation, support tickets, code, product usage data | VP Product, Head of Platform Data, General Counsel | Customer contracts often restrict reuse |
| Telecom, retail & industrial | Call transcripts, sensor and telemetry, catalogs, operations logs | Head of Data & Analytics, Digital Transformation lead | Employee and customer PII in free text |
| Legal, professional & research services | Contracts, filings, expert-labeled documents, domain corpora | Knowledge Management lead, General Counsel, Head of Innovation | Client confidentiality and privilege |
Illustrative, not exhaustive. Whether your data is licensable depends on your rights and contracts.
The systems where valuable data already lives
AI developers want records that show how real work gets done. They come from the tools your teams already run on, and we work from structured, de-identified exports, not access to your live systems.
Data warehouses
BigQuery, Snowflake, Databricks
Structured tables and schemas that mirror how the business runs
Chat & messaging
Slack, Teams, Google Chat
Conversation threads with participants and reply structure
Email & calendar
Gmail, Outlook
Threads and scheduling context between people and teams
Documents & files
Drive, SharePoint, Office
Contents plus authorship, timestamps, and metadata
Work & knowledge
Jira, Confluence, Notion, Asana
Tickets, wikis, and project history
Business platforms
Salesforce, HubSpot, Zendesk, ServiceNow, SAP
CRM, support, HR, and ERP records
You decide what leaves your walls
Governance is built into the deal structure, not added at the end. These are the terms we build into every agreement.
What you contribute
- De-identified records and workflows from the systems you choose
- Your sign-off on every data package before it ships
What never leaves
- Employee and customer identities
- Raw systems, credentials, or source access
- Anything outside your signed agreement
Is your company a fit?
AI labs generally prefer established operators, often 30 to 50 employees or more, though we assess every situation individually. Beyond size, they look for:
Established data systems
Years of records in the tools your teams already use.
Real operational workflows
Day-to-day work, not demos or synthetic data.
Domain depth
Specialized knowledge that makes your data hard to replicate.
A clear internal owner
Someone who can scope the deal, approve data packages, and keep legal and security reviews moving.
Questions companies ask
Is my data actually valuable?
Sometimes, and sometimes not. Value depends on uniqueness, quality, rights clarity, and lab demand. The assessment is designed to give you an honest answer, including when the answer is that the data isn't worth licensing.
Do I lose ownership of my data?
Not necessarily. Most deals are licenses, not sales. Scope (training only vs. redistribution), exclusivity, term, and deletion rights are negotiated terms, and they are where much of the value is decided.
What happens to the personal data in my dataset?
It is removed or de-identified before an AI lab sees anything beyond a summary. Scrubbing is verified by sampling, and we document the method so your legal and compliance teams can review it. No scrubbing process is perfect, which is why contract protections matter as well.
How is Verilex paid?
Verilex may be compensated by AI labs or their data partners when a license closes. We confirm all compensation terms in writing before you engage us.
Will AI companies use my data to compete with me?
Most labs want to improve how models handle roles like marketers, analysts, or support agents, not to replicate your business. Field-of-use limits, exclusivity terms, and competitor carve-outs are negotiable, and we build them into the agreement where your business needs them.
Do I have to hand over everything at once?
No. Deals can start with a narrow scope, such as one system, one team, or a time-limited slice, and expand once both sides are comfortable. We'll flag it whenever a structure trades flexibility for price.
Who does the anonymization?
Our approach is to de-identify data inside your environment before anything is extracted, so raw records never leave your systems. The method is documented and sampled so your security and legal teams can verify it.
Can you guarantee a sale or a price?
No. Anyone who guarantees a price for your data before reviewing it is selling you something else. We guarantee a clear assessment and a transparent process.
Start with a conversation
Tell us what kind of data you hold and roughly how much. We'll tell you whether it's worth taking to market.
Please don't send raw datasets or personal data by email. We'll set up a secure channel.