Selling meeting recordings and transcripts for AI training

A company can license its meeting recordings and transcripts to AI buyers only as far as the consent behind each recording reaches, and that consent was given to record a meeting, not to train a model. Federal law lets one participant record a call " where such person is a party to the communication", but California, Florida, Maryland, New Hampshire, Pennsylvania and Washington require every party ( Cal. Penal Code 632.7, Fla. Stat. 934.03(2)(d), Md. Code 10-402(c)(3), N.H. 570-A:2, 18 Pa.C.S. 5704(4), RCW 9.73.030(1)), and two states go further for business recordings: Illinois forbids passing telemarketing recordings made on one party's consent to " any third party", and Pennsylvania lets a customer-service business use its recordings " only ... for the purpose of training or quality control" and destroy them within a year. What buyers value in meetings is what no recording studio can stage: real decisions being made, several people talking over one another, " pauses, interruptions, overlapping speech and natural response", and transcripts aligned to the audio. But a voice is personal data, and in Illinois a " voiceprint" is a biometric identifier that no business may " sell, lease, trade, or otherwise profit from". So most meeting archives sell as de-identified transcripts, and the audio sells only where every speaker agreed to this use.

This guide is for a founder, officer or trustee (a holder, in our words) of a company that recorded its meetings in Zoom, Google Meet or Microsoft Teams, or ran a notetaker that joined calls and wrote transcripts. It covers what a meeting archive holds and where it sits, what each vendor's terms say about getting it out and training on it, recording consent state by state for internal meetings, customer calls, sales calls and interviews, the employees and customers whose voices are inside, de-identifying voices and transcripts, what meeting audio and transcripts fetch on each side of the market, preparing and delivering an archive, a closed company's case, and how a sale through AIData.Broker works. Every statute links to its own text and every number to its record.

AIData.Broker keeps the price board, the rates page and the audio and speech page this guide draws on, and it acts as the agent of holders who license their data, so we are a party in this market. Check any figure you plan to rely on against its source.

This guide is research from public sources and the statutes themselves, not legal advice.

Contents

1. What a meeting archive holds, and what buyers value in it
2. Where the recordings sit, and what each vendor's terms allow
3. Recording consent, state by state
4. Four kinds of meeting, four consent questions
5. Employees and customers: the people in the recordings
6. De-identifying voices and transcripts
7. What meeting audio and transcripts fetch
8. Preparing and delivering a meeting archive
9. A closed company's recordings
10. How a sale through AIData.Broker works
11. Questions people ask
12. What this means for you

What a meeting archive holds, and what buyers value in it

A company that has recorded its meetings for a few years holds three different things, and buyers treat them differently.

The recordings themselves. Video and audio of internal meetings (stand-ups, planning, reviews, all-hands), calls with customers (onboarding, support escalations, account reviews), sales calls (discovery, demos, negotiations) and interviews (candidates, user research, expert calls). Each recording carries the voices of everyone present, usually their faces too, and often a shared screen with documents, dashboards or code on it.

The transcripts. Text written from the audio, by the meeting software or by a notetaker, usually with a speaker label and a time on each line. A transcript is the part of a meeting that can travel without the voices, which is why it is the part most archives can actually sell.

What surrounds them. Meeting chat, shared files, calendar entries naming who was invited, and notes or summaries written afterwards. These tie a meeting to the decisions, tickets and documents that followed it.

What buyers value

Buyers of company data pay for the working record of how a business actually runs (our page on what AI labs buy sets out each kind), and meetings are where much of it is decided out loud. Three qualities set meeting material apart from other speech and other text.

Real decisions. A meeting is people with different information and different stakes arriving at a decision: a price, a hire, a design trade-off, a fix. That is the material agents and assistants are now trained to follow, and it rarely exists in written form anywhere else. Buyers who price whole company archives name the communication systems first: micro1's chief revenue officer said, on the record, that without Slack, CRM or email records the rest of a corpus is worth less, and the same record of price drivers says value rises with the number of white-collar staff ( what raises a corpus price). Records that connect to each other (a meeting tied to the ticket or document that came out of it) sell above loose files ( linked records sell above loose documents).

Multi-speaker conversation. Most speech data is one person reading, or two people chatting to order. Real meetings have several speakers, interruptions, cross-talk, people correcting themselves and others, and specialist vocabulary. Voice-agent builders want exactly that: Ocular AI, a voice-data start-up reported to sell to frontier labs, is described as supplying " material covering pauses, interruptions, overlapping speech and natural response". Collectors pay their own recorders more for conversation than for reading: $28 an hour for an unscripted two-person conversation at Perit AI, against a base of $20 per approved audio hour for read speech.

Transcripts aligned to audio. Audio with an accurate transcript, each line tied to its speaker and its time, is worth far more than either alone, because a recognition model learns from the pair. Datasets of real support calls that are sold with verbatim transcripts and speaker separation ask $24,500 for 1,000 hours, against well under a dollar an hour for bulk call audio with neither ( Hindi call audio, as listed).

Sales calls in particular are now bought. In one purchase a lab buyer took a company's corpus that included about 83K sales-call recordings beside its code, tickets and documents. Workspace buyers price meetings as part of the whole: Corpus prices a company's workspace data, which our record of its offer describes as email, chat, meetings and CRM, at engagements from $100,000 to $2 million, and Polyshares states $100K to $2M and above for a company's operating record.

Where the recordings sit, and what each vendor's terms allow

Where a recording sits decides who can export it and on what terms. The three meeting services most companies use each have terms that say who owns the content and how it may leave. We read them for our rights pages; the quotes below come from those readings.

Zoom

Zoom's terms leave ownership with the customer and say that Zoom itself does not use " audio, video, chat, screen sharing, attachments or other communications-like Customer Content" to train its own or third-party models. So, on Zoom's own terms, it has not trained its models on your meetings, which matters when a buyer asks whether the content is exclusive (a question buyers' counsel ask of every system: has a vendor already trained on it).

The restriction that matters is in Zoom's API terms. Content " accessed or obtained using the Zoom APIs or your Application" may not be used for AI training except with Zoom's express written permission, and the one exception, for an account that consents, allows a model " solely for AI Training specifically for their use": the resulting model may be used only by that customer. A buyer's connector, a labelling vendor's app or any script that pulls recordings through Zoom's API to build a model for others falls inside that bar. The route Zoom's terms leave open is the one its own records describe: hosts and administrators downloading recordings and transcripts themselves. Our reading of the Zoom terms, risk rating and notes are on the Zoom rights page, which treats recordings as the content with "the highest consent bar".

Google Meet

Google's terms for Workspace draw the same line between an administrator's own export and an app's access. The organisation-wide export is open to every edition and is run by a super administrator whose account is " at least 30 days old", and Google says " The export is available no earlier than 48 hours after you start the export". Apps that reach Workspace data through Google's APIs are barred from " Transferring or selling user data to third parties like advertising platforms, data brokers, or any information resellers" and from training general models on it. Wherever Meet recordings and transcripts sit in your Workspace, then, the route out is the native export or Vault, run by your own administrator, with the files passed to the buyer under your own licence.

Microsoft Teams

Microsoft's data protection terms state the customer's export right most explicitly of the three, during the subscription and, Microsoft says, " in a limited function account for 90 days after expiration or termination". Bulk export of Teams content for an organisation usually runs through Microsoft Purview eDiscovery, and Microsoft's own documentation ties premium eDiscovery features to E5 licences. Microsoft's API terms forbid an app to " Scrape, build databases or otherwise create copies of any data accessed or obtained using the Microsoft APIs" beyond its intended use, and a Product Terms change of May 2026 says that " outputs generated through Robotic Process Automation, bots, or other similar technologies may not be used to develop, train, evaluate, improve AI or machine learning models". That clause is written for the Office applications, and our reading of it is that it does not on its face stop a customer licensing its own content; whether it reaches a transcript a bot produced inside Teams is not settled by anything we read, so the careful course is to export the recordings and transcripts themselves through your own administrator, not the outputs of a bot.

Notetakers

A notetaker is a service that joins a meeting as a participant, records it and writes a transcript and summary, keeping copies in its own systems. We have not read any notetaker's terms for our rights pages, so this guide cannot say what any of them allows. The questions to put to each are the same as for the three services above: who owns the recordings and transcripts, how an administrator exports all of them, and whether its terms restrict training on what leaves it. Two points hold whatever the terms say. A notetaker's copy is still a recording of the meeting, so the consent questions below apply to it in full. And if the notetaker was connected through an individual employee's account rather than the company's, the company may not be able to export those recordings at all without that employee.

The export route matters as much as the right to sell

Across all three services, the pattern is the same: the customer owns the content and can export it, and the vendor's API terms restrict what an outside app may do with content it pulls. Buyers' counsel ask how the data left each system, and the answer they want is that " Each system was exported by its own admin with the vendor's native export tool". So never let a buyer or a broker connect an app to your meeting account to pull recordings for training. Export them yourself, and record how.

Recording consent, state by state

Before anything about AI, a recording has to have been lawfully made, because a defect in how it was made follows it into every sale. The federal Wiretap Act is a one-party rule: it is not unlawful for a person " to intercept a wire, oral, or electronic communication where such person is a party to the communication or where one of the parties to the communication has given prior consent", unless the purpose is criminal or tortious. A company whose own staff hosted and recorded the meeting is a party, so federal law is satisfied by its consent. But the same Act separately punishes anyone who intentionally discloses or uses the contents of a communication knowing it was unlawfully intercepted, and several states are stricter than the federal floor.

The table covers the states whose statutes, read in their own text for our audio research, depart from the one-party rule. "All parties" means every person on the call or in the conversation must consent. States not in the table were not read. A video meeting is a call for some purposes and a conversation for others; the statutes we read speak of phone calls and oral conversations, so a meeting with people dialled in from several states is safest read under the stricter column.

Four points from the table matter most for meetings.

California's mobile-phone rule has no confidentiality condition. Its general section reaches only a confidential communication, which excludes a conversation where the parties may reasonably expect to be recorded. But its section on cellular and cordless phones forbids recording a call " without the consent of all of the parties to a communication" with no such exception. Anyone who joined a meeting by phone from California is covered by the stricter rule.

Notice works differently in each state, and it has to be on the recording. In Washington, " consent shall be considered obtained whenever one party has announced to all other parties" that the conversation is about to be recorded, if the announcement is itself recorded. Connecticut accepts a verbal notice recorded at the start or " an automatic tone warning device". In Montana, " If one person provides the warning, either party may record". In Illinois and Massachusetts the offence is secret recording: Illinois defines surreptitious as " obtained or made by stealth or deception, or executed through secrecy or concealment", and Massachusetts reaches a person who would " secretly hear, secretly record". A meeting where everyone saw that recording was on is in a better position than a call recorded silently, but a buyer asks for proof per batch, not a policy saying notices were shown.

Oregon is stricter in the room than on the phone. A participant may record an Oregon phone call, but an in-person conversation may be recorded only if " all participants in the conversation are specifically informed". A meeting room in an Oregon office, recorded by a device on the table, is held to the stricter rule.

A meeting has many ends. A call with participants in Texas, California and Washington involves all three states' rules, and a meeting archive rarely records where each person sat. The careful reading is the strictest state any participant could have been in. In practice that means sorting meetings by who attended (staff by office, outsiders by the address on their account or company) before a buyer sees them.

Consent to record is not consent to train

Every statute in the table governs the act of recording, and several govern using or disclosing a recording that was unlawfully made. None of them, as read for this guide, says whether a recording that was lawfully made may later be licensed to train a model without asking the participants again. No statute we read answers that question either way. Two states do speak to later use, and both bear on business recordings.

Illinois forbids passing on telemarketing and opinion-research recordings. A business engaged in marketing, opinion research or telephone solicitation may record calls on one party's consent for quality control, staff training and internal research, but nothing obtained that way may be " divulged to any third party". Licensing such recordings to an AI buyer divulges them.

Pennsylvania limits customer-service recordings to the business's own use, and to one year. " Any communications recorded pursuant to this paragraph may only be used by the business for the purpose of training or quality control", and unless other law requires otherwise they "shall be destroyed within one year from the date of recording". Licensing them to an outside buyer is use by someone other than the business.

The defect travels to the buyer. The federal Act punishes using or disclosing contents known to be unlawfully intercepted; Oregon bars obtaining a recording " while knowing or having good reason to believe" it was unlawfully made; Illinois makes it eavesdropping to use or disclose information one knows or reasonably should know was obtained in violation of the Article. This is why a well-advised buyer asks for consent records meeting by meeting, and will not take an archive whose consent basis nobody can show.

Outside the US the answer is clearer. For holders in Canada, the federal privacy law says that when personal information " is to be used for a purpose not previously identified", consent is required before it is used that way, and the Privacy Commissioner's guidance on call recording says an organisation " may only use the information collected for the purposes specified". For holders in the UK, the Information Commissioner's guidance says a very different or unexpected purpose is likely to be incompatible and that " you're likely to need to ask for specific consent". Licensing identifiable meeting recordings for AI training is the kind of new purpose both describe.

Four kinds of meeting, four consent questions

Meetings are not one category. The people in them, what they were told and what they said differ by kind, and so does what can be sold. Buyers' counsel ask whether calls and meetings were recorded with every participant's consent, and whether that consent covered use beyond the meeting; the answer they want is " Recording notices on every call, written consent from internal participants for AI licensing, external participants (clients, candidates) excluded or separately consented". The red flag on the same record is " meetings with external guests".

Internal meetings

These are the easiest to clear, because every participant works for the company and can be asked. They are also the richest record of how decisions get made. Two things still need care. First, the notices: New York requires an employer that monitors employee email to give acknowledged written notice on hiring, and Connecticut requires prior written notice of the types of electronic monitoring; our employees guide covers these and the other monitoring laws. Second, what is said: internal meetings include pay discussions, performance reviews, HR complaints and legal advice, all of which come out before anything else is considered (section 5).

Customer calls

A customer call adds two problems at once. The customer's people are outside participants who consented, at most, to being recorded, and many of them may have joined from all-party states. And what they said is often the customer's confidential information, governed by the customer contract. Buyers' counsel ask whether your customer contracts and data processing agreements allow using it for anything other than serving them, and whether the data contains other companies' confidential information. Customer calls usually sell, if at all, as transcripts with both the people and the customer's business removed.

Sales calls

Sales calls are what buyers ask for by name, because they are task-oriented conversations with objections, questions and outcomes. They are also the most likely to have been recorded under the thinnest notice, and to include people from every state. Treat them as customer calls with one extra step: if the recording tool kept no record of who joined from where, sort by the address on the prospect's company record before deciding which calls can go forward.

Interviews

Candidate interviews carry personal history and are covered by employment law beyond recording consent; they are usually better left out. User-research and expert interviews are different: participants often signed a consent form, and if that form already covers further use, research interviews can be among the cleanest audio a company holds. Read the form. If it says the recording is "for internal research only", it does not cover a licence.

Employees and customers: the people in the recordings

A meeting archive belongs to the company as a matter of contract with the vendor, but the voices, faces and words in it belong to people, and they decide what can be sold.

Employees in California. California treats staff data as personal information. Buyers' counsel ask whether current and former staff received notice and an opt-out before the sale, or whether employee personal information was removed to the deidentified standard. That standard needs all three of reasonable technical measures, a public commitment not to re-identify, and contracts binding every recipient ( Cal. Civ. Code 1798.140(m)).

Pay, conditions and unions. Meetings where staff discuss pay, working conditions or union matters are a separate exclusion: buyers' counsel want union and HR channels and any discussion of pay or conditions left out, and a collective agreement checked for data and monitoring clauses.

Staff in the EU or UK. For staff there, the question is the legal basis that lets you reuse the personal data in their work (messages, documents, code history, recordings) for a sale; only data anonymised to the EU standard falls outside the GDPR.

Biometrics. A voice can be biometric data. Illinois lists a " voiceprint" and "scan of hand or face geometry" among biometric identifiers, forbids any private entity to profit from them, and lets each person sue for at least $1,000 per violation, or more if intentional. Meeting video adds faces to voices. Whether raw audio or video, as opposed to a voiceprint or face template extracted from it, counts as a biometric identifier is contested in pending cases, which is why our rights check on biometrics treats call-centre recordings, meeting videos as a red flag and asks for written releases naming AI training and sale, Illinois residents excluded or separately released, and a buyer who promises not to extract identifiers.

What your policies said at the time. Buyers ask what your privacy policy, terms of service and employee handbook said about sharing or selling data at the time it was collected. A handbook that said meetings may be recorded "for internal purposes" is a promise about use, and a licence goes beyond it.

Legal advice and secrets. Meetings with outside counsel, board sessions discussing litigation and calls where trade secrets were walked through come out, because de-identification does not cure privilege ( legal advice, litigation material, or trade secrets).

The order of work that follows is the one our employees guide sets out for all internal communications: confirm title, check notices, exclude, de-identify, write it down. For meetings, recordings with outside participants go first into the "exclude unless consented" pile.

De-identifying voices and transcripts

Text can be de-identified well; voices cannot yet. That difference decides how most meeting archives can be sold.

A voice identifies its speaker, and disguising it does not reliably hold. Voice anonymisation is an active research field, and its own benchmark is sobering: the organisers of the VoicePrivacy 2024 Challenge reported that systems which first appeared to hide speakers well were later shown to be vulnerable to stronger attacks. For that reason our de-identification standard does not sell recordings of voices unless the people in them consented in writing to AI training and sale; transcripts are rewritten instead. Buyers draw the same line from the other side: micro1's licence terms have the holder warrant that the data contains no biometric identifiers or biometric information unless the order form says otherwise, and HUMXN says " Names, faces, voices, account numbers and other personal data are removed or masked" before buyers see a dataset. Meeting video is harder still, because a face is on screen for the whole call and shared screens show names, email addresses and documents.

Transcripts are where most meeting archives can be sold. A transcript carries what a meeting was about (the problem, the options, the decision, the way people actually talk at work) without the voice. Rewritten so that no person can be identified, it is a text record of the same kind as a Slack export, and the same buyers want it: Handshake's outbound team licenses companies' code and support transcripts among their operating data. Meeting transcripts need more than name replacement, because speakers name each other constantly ("as Priya said last week"), mention customers and contracts by name, and refer to things only one person could know. The engine behind our standard was built for this kind of record:

• It runs on your computer. The engine works in your browser, so nothing identifying leaves your computer: not the transcripts, not the list of people it builds, not the key.
• It protects a person in every mention. A speaker counts as protected only when every mention of them is replaced, with each person's name, first name, surname and handle linked to one stand-in, so the speaker labels and the names spoken in the meeting agree.
• It screens content as well as names. Meetings about health, children, beliefs or crime are flagged under its content screen, because removing names does not make a health conversation safe to sell.
• A person reads a sample. Rules read every record, and a person reads a random sample of the rewritten ones before anything goes further.
• The buyer commits in writing. Every licence binds the buyer not to re-identify the data or link it to other data to do so.

On company Slack, email and documents the engine scores 86.8% on a public benchmark, and we call its output de-identified, with the measured risk, not anonymous. You can run it yourself, free, with the de-identification tool. For US staff and customer data, buyers ask whether the result meets the CCPA definition, including a public commitment and contract terms binding the buyer.

Where every speaker consented, spoken details still come out. Audio sold as audio still carries names, account numbers and addresses spoken aloud. Those spans can be silenced or replaced, and the datasheet should say which method was used. Datasets of real support calls on sale are offered PII-redacted for that reason.

What meeting audio and transcripts fetch

Meeting material is priced two ways, and holders should know both. Audio is priced per hour of speech, like any speech data. Transcripts and the rest of the meeting record are priced as part of a company's working record, by the company, not by the hour.

Audio, per hour, on each side of the market

The same hour of speech carries very different prices depending on where in the market it is quoted. The rates page computes each side from every public price at once.

Meeting audio specifically has few public prices. The clearest is a research corpus: the ICSI Meeting Speech corpus, about 72 hours of English meeting speech, licenses to non-members for $1,900, and that licence excludes commercial use. The nearest commercial comparisons are conversational: transcribed two-person conversation in US English lists at $90 an hour, and real support calls with verbatim transcripts at $24,500 for 1,000 hours or $25 an hour usage-based. Those are asking prices for audio whose consent chain is documented and whose speakers are separated. A company's meeting archive, recorded for the minutes and mixed on one track, starts lower, for the reasons our audio guide sets out: the consent was given for something else, it was not recorded for a model, and it is full of personal details.

The chart puts the figures that bear on meeting audio on one axis.

Sources: read speech, conversation, real support calls, two-person conversation.

Transcripts and the meeting record, priced with the company

Meeting transcripts rarely sell alone. They sell as part of a company's working record, beside its chat, email, tickets, CRM and documents, and those records are priced per company. The median of the per-company sales on record is $28.3K per company, and the middle half of public prices for company records runs $174K to $693K per company. Buyers that price workspaces publish ranges that name meetings among the systems: Corpus at $100,000 to $2 million per engagement for email, chat, meetings and CRM, and Polyshares at $100K to $2M and above. Systems multiply each other: Ooak's estimator values the same company several times higher with six connected tools than with three ( connected systems multiply value). A meeting archive's main contribution to a price is often that it connects the rest: the call where the decision was made, tied to the ticket and the document that followed.

What moves the price

• Up: every speaker's written consent to AI use (the only route to selling audio as audio); original files with each speaker on a separate track; accurate transcripts aligned to the audio, with speaker labels; specialist subjects and less common languages or accents; meetings linked to the records that followed them; years of history across many teams.
• Down: one mixed track, compressed copies, no transcripts; outside participants with no consent beyond recording; health, pay or legal content scattered through; recordings from all-party states with no proof of notice; a buyer who demands exclusivity without paying for it (an exclusive licence fetches about 2.5 to 5 times a non-exclusive one, and non-exclusive data can be sold again).

Preparing and delivering a meeting archive

Preparing a meeting archive is mostly sorting it by who was in each meeting and what they agreed to, then making the part that can be sold as useful as possible. A buyer's lawyers ask 27 rights questions of a data sale, and 16 of them stop it until answered; for meetings, most of the blockers are about consent and the people inside. The order below is the one to work in.

1. Inventory what you hold, system by system

For each service (Zoom, Meet, Teams, any notetaker), list the number of recordings, the hours, the date range, which have transcripts, which have separate audio tracks, and who can export them. Note recordings kept on individual laptops (Zoom's local recordings sit on hosts' computers) and notetakers connected through personal accounts, because the company may not be able to reach either without the person.

2. Classify each meeting by who was in it

Use the calendar entry or the participant list to sort every recording into internal, customer, sales or interview. Mark which included anyone outside the company, and where each participant was likely to be, using staff offices and the addresses on outside participants' company records.

3. Map consent, meeting by meeting

For each batch, record what notice was given (a banner, a spoken announcement, an invitation that said recording was on) and whether it is on the recording itself, what your handbook and privacy notice said at the time, and whether any participant signed a consent that names AI training. Then sort into four lots:

• Every speaker agreed to AI use. Internal meetings where staff signed a consent that covers licensing for AI training; research interviews whose form covers further use. This audio can be sold as audio.
• Lawfully recorded, no AI consent. Most meeting archives. Sell transcripts, de-identified, or ask the participants; a consented slice of recent internal meetings is often enough for a first sale.
• Doubtful or defective. Meetings with participants in all-party states and no proof of notice, telemarketing calls recorded in Illinois under the one-party exemption, customer-service calls recorded in Pennsylvania under its exception, anything under legal hold. Leave these out.
• Not yours to grant. Calls recorded on a customer's own meeting account, or a partner's, and notetaker copies held under someone else's account.

4. Exclude by content

Before any de-identification, take out meetings with outside counsel, board and investor sessions, HR and performance conversations, pay and union discussions, health matters, and calls where customers walked through their own confidential information. A shared screen can carry a customer's dashboard or a competitor's document, so video needs a separate look even where the audio is clean.

5. De-identify the transcripts

Run the transcripts through the de-identification tool, in your browser, with the speaker labels included so that each person's stand-in is the same in the label and in the speech. Check the summaries and meeting chat too, which name people more freely than the speech does.

6. Describe it

A buyer cannot listen to thousands of hours. A datasheet that states hours and meetings by kind, speaker counts, languages, recording conditions, transcript accuracy, consent basis per lot and what was removed is what lets it say yes quickly; the datasheet tool structures one. When you add files to the valuation, your browser reads them: recordings are counted, their words are read from transcripts (VTT and SRT files), and what leaves your computer is counts, with no text and no names of people.

7. Deliver it yourself

Export with each service's own admin tool, keep a note of how and when, and deliver the files to the buyer yourself or through a provider you choose. Never give a buyer access to your meeting account, and never authorise its app to pull recordings: Zoom's and Google's API terms bar exactly that use, and Microsoft's bar building copies.

Example: a small software company. Take a company with four years of Zoom cloud recordings: weekly planning and design reviews, customer onboarding calls, and sales demos, with transcripts on most of them, staff in California, New York and Texas, customers everywhere, and a handbook that says meetings "may be recorded for internal use". Its first finding is that its internal meetings are its strongest material and the easiest to clear: staff can be asked to sign a consent naming AI licensing, the New York staff were given their monitoring notice, and the transcripts, de-identified, are a record of how a software team makes decisions. Its second is that the customer and sales calls are harder: participants from California joined by phone, under a rule that needs all parties, customers' own systems are discussed on shared screens, and the customer contracts limit use of customer information to serving the customer. Its practical route is a lot of de-identified internal transcripts, with the audio of staff who consented, beside its Slack, tickets and code; customer and sales calls stay out, or go in as transcripts only after the customer's business has been removed. It exports through Zoom's own download route, not an app.

Example: a recruiting firm. Take a firm whose recruiters recorded every candidate screen and every intake call with the companies it hires for through a notetaker connected to each recruiter's own account. Its archive looks large but most of it cannot be sold: candidates discussed pay, health and personal history, intake calls carry those companies' confidential information, and the notetaker copies sit under recruiters' personal accounts, some of whom have left. Its saleable material is narrow: internal team meetings, transcripts de-identified, if the firm can export them.

A closed company's recordings

A closed company often holds the most complete meeting archive, because nobody deleted anything, and the least time to sell it. Three questions decide whether its recordings can still be sold.

Do the recordings still exist? Meeting services keep recordings while the account is paid for. Microsoft keeps customer data in a limited account " for 90 days after expiration or termination" so the customer may extract it. We hold no comparable reading for Zoom's or Google's retention after an account lapses, so assume the recordings go when the subscription does: keep the account paid, or run the full export before it lapses. The deletion clock shows the dates we hold for each system.

Who can sign? The company sells, not its founders personally, and the right person signs for it. Buyers' counsel ask for the legal entity that collected the data to be the licensor, with the board's approval. A dissolved Delaware corporation continues for three years to wind up its affairs, including disposing of its property, and in a bankruptcy the trustee signs, with a sale that includes personal information running through section 363(b)(1) and its privacy conditions. Buyers ask that the right officeholder signs. Our estates page sets out the path for trustees, assignees and dissolved companies.

Who can run the export, and who can be asked? The administrator of the meeting account may have left; transfer the role to an officer or the trustee first. And the staff whose consent would let audio be sold are now harder to reach. That pushes most closed companies toward de-identified transcripts, with audio only from people who can still be asked. The employee question is what delays records sales most: in the Spirit Airlines records sale, the hearing has been adjourned repeatedly while the debtors negotiate with objectors over personally identifiable or sensitive employee information. Settle it before going to market.

How a sale through AIData.Broker works

AIData.Broker represents companies selling their data to AI buyers. We act only for the holder, as its agent in the sale, for holders based in the United States, the United Kingdom and Canada, and the holder itself licenses its recordings and transcripts to the buyer. We take on a lot only where the expected sale reaches $26,000; smaller archives are usually better served by approaching a buyer directly, and the page on going direct or through us sets out the trade-off.

The steps are the same as for any lot, set out in how it works:

1. A valuation first. The valuation puts a range on your records from public prices, binds no one and needs no upload. For speech it asks the hours, the kind of speech, the language and whether every speaker agreed; for the rest of the working record, it asks about your systems, years and staff.
2. The rights file. Together we sort the archive by kind of meeting and by consent as in the sections above, decide what is sold as audio and what as de-identified transcripts, and write the datasheet.
3. Every buyer that fits, at once. The lot goes first as a description that does not name the holder to every buyer whose needs and terms fit and that we can screen, and all of them answer by the same deadline. The holder may leave out any buyer it chooses.
4. Offers in writing, side by side. Every offer reaches the holder with its full terms as the buyer stated them and what the holder would keep after our fee. The holder alone decides whether to accept, which buyer and on what terms. Our free offer check places any written offer you already have against the market and lists that buyer's clauses.
5. Delivery and payment between holder and buyer. The holder, or a provider it chooses, delivers to the buyer through its own exports, and the buyer pays the holder directly. We do not buy your recordings or license them in our own name; we hold only what you choose to share with us to value and sell them, such as counts, the rights file and the datasheet.

Our fee is a success fee on what the holder is actually paid, band by band: 15% of the first $250,000, 10% from $250,000 to $1 million, and 6% above $1 million, as the pricing page sets out, so on a sale paying the holder $300,000 the fee is $42,500. Nothing is payable if nothing sells. In a sale we run we take nothing from any buyer: where a buyer offers an introduction fee, we decline it and ask the buyer to put it into the holder's price instead. The mandate is exclusive for 2 months, with a tail of 12 months for sales to buyers we introduced; the binding rules are in the mandate, which our terms describe.

Questions people ask

Can a company sell its Zoom recordings for AI training? Potentially, if the company exports them itself and every participant's consent covers the use. Zoom's terms leave ownership with the customer and let hosts and administrators download recordings, but its API terms bar using content obtained through its APIs to train, develop, evaluate, test, improve, or modify AI models without Zoom's permission, so no buyer's app should pull them. Most archives sell as de-identified transcripts, with audio only from participants who agreed in writing.

Are meeting transcripts easier to sell than the recordings? Yes. A transcript rewritten so that no person can be identified carries the meeting's content without the voice, which is personal and, in Illinois, can be biometric. The recording-consent question still applies to how the meeting was recorded, but a de-identified transcript removes the biometric and most of the privacy exposure. The de-identification tool rewrites transcripts in your browser.

Is "this meeting is being recorded" enough consent to sell the recording? No statute we read says it is, and none says it is not. A notice can satisfy the recording laws (in Washington a recorded announcement counts as everyone's consent), but it speaks to being recorded, not to later use, and in Canada and the UK a new purpose generally needs new consent ( PIPEDA, ICO). Buyers treat it as a gap and price it.

Which states require everyone's consent to record a meeting? From the statutes read for this guide: California, Florida, Maryland, New Hampshire, Pennsylvania and Washington for calls and conversations alike; Connecticut for calls, by its civil statute; Nevada for calls, in effect; Oregon for in-person conversations; Massachusetts and Illinois for secret recording; Montana requires the knowledge of all parties; Delaware and Michigan are disputed. The table above links each to its text.

Can we sell recorded sales calls? They are in demand: one lab buyer took a corpus with about 83K sales-call recordings. But prospects are outside participants, often in all-party states, who consented at most to recording. Sell transcripts de-identified, with the prospect's business removed, and audio only from calls whose participants agreed to this use. Illinois forbids passing on telemarketing recordings made on one party's consent.

What is an hour of meeting audio worth? Licensed speech sold for a median of $76 per hour of audio an hour and ready-made speech datasets ask $118 per hour of audio, but those are prices for consented, documented audio. A research meeting corpus licenses at $1,900 for about 72 hours, for research only. Meeting transcripts are usually priced with the company's whole working record, where the per-company median is $28.3K per company.

What about recordings from our notetaker? They are recordings of the meeting like any other, so every consent question applies. We have not read notetakers' terms, so check each one for ownership, admin export and any restriction on training, and confirm the company, not an individual employee, holds the account.

Can a closed company still sell its meeting recordings? Yes, if the recordings still exist and the right person signs: the officers during a Delaware corporation's three-year wind-up, or the trustee in a bankruptcy. Keep the meeting accounts paid, or export before they lapse.

What this means for you

Selling meeting material is a consent exercise with a price attached. The audio is worth most where every speaker agreed to this use, and that is rarely true of an archive recorded for the minutes. The transcripts, de-identified, are what most companies can sell, and their value comes from what they connect: the decisions behind the tickets, documents and code a buyer is already paying for.

1. Know your number before anyone quotes one. The rates page and the price board hold every public price per hour of audio and per company, and the valuation puts a range on your own records from them. The audio and speech page keeps the current prices, buyers and 11 lots of speech on offer.
2. Sort by meeting and by consent first. Classify every recording as internal, customer, sales or interview, map each to the states its participants were in and what they were told, using the state table, and split the archive into what can be sold as audio, what as de-identified transcripts, and what stays out.
3. Export it yourself. Use each service's own admin export, never a buyer's app, and record the route; the Zoom rights page and our rights pages hold each vendor's terms.
4. Sell the words where you cannot sell the voices. Rewrite transcripts with the de-identification tool, and read the de-identification standard for what buyers will rely on. Our guides to selling audio and speech, selling Slack data and employee emails and chats cover the rest of the record.
5. Let buyers compete. Put the lot before every buyer that fits at once, compare grants, exclusivity and withdrawal terms as well as price, and check any offer with the offer check. How it works sets out a sale through AIData.Broker step by step.