Yes, if the company owns the records and they describe its own infrastructure. Hosting and CDN providers, managed security providers, appliance makers, SaaS companies with a security team and ISPs hold incident tickets joined to the change that fixed them, configuration and infrastructure-as-code history, ordered logs, runbooks and postmortems. Secrets, client data and personal data come out first. A sample of two or three incidents is the usual first step.
- The valuable unit is an incident joined to its fix: ticket, chat thread, config diff or pull request, and outcome.
- Records about your own infrastructure are the candidate. Records from a client's environment need that client's agreement.
- Secrets in config history and Terraform state come out before anything else.
- Buyers start with a small sample, not a full export.
Can a company license its incident and configuration records to AI?
Yes, when two things are true: the company owns the records, and the records describe infrastructure the company runs and defends itself. Incident tickets, configuration history, logs of what happened, runbooks and postmortems about your own systems are yours to license once secrets and personal data are removed. Records from a client's environment are a different case and need that client's agreement.
The answer changes record by record, so sort them before anyone talks about an export. The table shows the usual split. This page is general information, not legal advice.
| Record | Usually licensable? | Main condition |
|---|---|---|
| Incident tickets and postmortems about your own platform | Yes | Remove customer names, staff personal data and secrets |
| Config and infrastructure-as-code history (Terraform, Ansible, firewall rules, routing) | Yes | Scan and strip credentials; Terraform state needs extra care |
| Runbooks, on-call records and SOC tickets for your own systems | Yes | Remove phone numbers and personal schedules |
| Ordered logs or captures of an attack on your own systems | Often, as part of an incident record | Anonymize addresses and drop payloads first |
| Tickets, logs or alerts from a client's environment | Only with the client's agreement | Read the services contract and data processing terms |
| Subscriber traffic, customer content, customer logs | No, as a rule | These belong to, or are protected for, the customer |
Which companies hold infrastructure incident records?
Five kinds of company hold these records in volume: hosting and CDN providers, managed security and SOC providers, network and security appliance makers, SaaS companies with their own security team, and ISPs. Each runs real infrastructure under attack, changes it often and writes down what it did.
| Who | What it usually holds | Where the line falls |
|---|---|---|
| Hosting, cloud, colocation and CDN providers | Platform incidents, WAF and firewall changes, abuse desk and DDoS records | Own platform records versus customer content and customer logs |
| Managed security, MSSP, MDR and SOC providers | Alert triage, investigations, SOC tickets, playbooks | Most cases come from client environments |
| Network and security appliance makers | Support cases, firmware and config history, their own corporate network | Own product and network versus device data from customers |
| SaaS companies with a security team | Terraform history, paging-tool incidents, postmortems, cloud audit logs | Own production systems versus customer data inside them |
| ISPs, WISPs and telecom operators | NOC tickets, routing changes, outage records | Network engineering versus subscriber data and CPNI |
What makes up a complete infrastructure incident record?
A complete record has five parts that can be joined by time and by ID. NIST defines an incident as "an occurrence that actually or imminently jeopardizes, without lawful authority, the integrity, confidentiality, or availability of information or an information system" (NIST SP 800-61r3). Outages that are not attacks leave the same trail: an alert, a ticket, a fix and a review.
| Record group | Examples | How it joins the others |
|---|---|---|
| Incidents joined to their fix | The ticket, the chat thread, the config diff or pull request, the outcome | Ticket ID in the commit message or pull request |
| Config and infrastructure-as-code history | Terraform, Ansible, firewall rule sets, routing configs, with commits and reviews | Commit time and author against the incident window |
| Ordered logs or captures | Firewall, WAF, DNS and proxy logs, packet captures of the attack and the response | Timestamps against the ticket timeline |
| Operations records | Runbooks, SOC tickets, on-call records, postmortems from the same period | The runbook used, the person paged, the review written |
| Edge rules | WAF, DNS and load-balancer changes made during an attack | Change time against the attack timeline |
Does it matter whose infrastructure the records describe?
Yes, it is the first question to answer. Records about infrastructure the company runs for itself are the company's own. Records produced inside a client's environment, such as an MSSP's alerts from a client network or a hosting provider's copy of a customer's logs, are usually covered by the client contract, and often by data protection law as well.
Under the GDPR, the controller is the body that "determines the purposes and means of the processing of personal data", and a processor "processes personal data on behalf of the controller" (GDPR Article 4). A processor that starts deciding purposes for itself "shall be considered to be a controller in respect of that processing" (GDPR Article 28). Licensing client-derived logs for a new purpose is exactly that kind of decision, so it needs the client's agreement first.
Are a SaaS company's Terraform history, incidents and postmortems worth licensing?
They can be, because together they show how a real production system was built, broke and was repaired over years. A SaaS company with a security team usually holds all three: infrastructure-as-code history in git, incidents in a paging tool, and postmortems that explain the fix. The value is in the join.
Paging tools already keep the order of events. PagerDuty describes its post-incident reviews as helping teams "build chronological incident narratives, and coordinate preventative follow-up actions" (PagerDuty docs). That narrative, the alerts behind it and the pull request that closed it form one incident record.
Terraform history needs the most care. HashiCorp warns that "Terraform state and plan files contain detailed information about your infrastructure, including resource attributes and metadata that can contain sensitive values, such as initial database passwords or API tokens", and tells teams to keep state out of git (Terraform docs). State files and secrets come out first; the guide to removing secrets from git history covers how.
What can a network or security appliance maker license?
Mainly records about its own product and its own network. A firewall, router or security appliance maker usually holds support cases that show how a configuration failed and how it was fixed, the version history of default configs and rule sets it ships, its own lab and test records, and the incidents and changes of its corporate infrastructure.
Device data sent back from customer sites is different. Telemetry, crash dumps and config backups uploaded by customers can carry their addresses, rule sets and credentials. Treat them like client records: they need a contract right or the customer's consent.
What has to come out before records leave the company?
Five things come out before a sample is cut: secrets, client and customer data, personal data, records under a legal or investigation hold, and details of weaknesses that are still open.
| Remove | Why | How |
|---|---|---|
| Keys, passwords, tokens, Terraform state | They give access to live systems | Scan and rewrite history; rotate first (guide) |
| Client and customer data | It is usually not yours to license | Drop client-environment records; replace tenant IDs |
| Personal data: staff names, phone numbers, IP addresses of users | Privacy law and staff expectations | Consistent pseudonyms (guide) |
| Evidence under a legal or investigation hold | It may be needed in court or by an insurer | Leave it out until the hold ends |
| Open weaknesses and unpatched systems | They tell an attacker where to look | Hold back recent incidents until fixed |
NIST tells incident responders to "collect and retain evidence from an incident in accordance with the organization's evidence preservation procedures and data retention policies" (NIST SP 800-61r3). Read those procedures before choosing which incidents go into a sample.
Example: a SaaS company sorts five years of records
A fictional SaaS company with 300 staff and a six-person security team wants to know what it could license. Its security lead sorts the records into three piles before anything is exported.
| Pile | Records | Decision |
|---|---|---|
| Never | Customer database snapshots, a forensic report written for outside counsel, an incident still being fixed | Excluded |
| Only after work | Terraform repository with 5 years of commits, paging-tool incidents, chat channels from incident calls, WAF logs | Secrets scanned and history rewritten, state files dropped, names and tenant IDs replaced, addresses anonymized |
| Ready first | Runbooks, 40 written postmortems, change tickets linked to pull requests | Reviewed for customer names, then three incidents picked and joined end to end |
Fictional example written for this page, not a real company's records. What the team wrote about its own systems moves first.
Where each part of the record is covered
Each guide below covers one part of the record or one kind of holder in depth.
| Topic | Guide |
|---|---|
| Hosting, cloud, colocation and CDN providers | Can a hosting or CDN provider license incident data? |
| ISPs and telecom operators | Can an ISP license network operations records? |
| Runbooks and playbooks | Runbook vs playbook |
| Postmortems | Incident postmortem examples |
| Joining ticket, chat, diff and outcome | Incident timeline template |
| Terraform and Ansible history | Infrastructure as code examples |
| Firewall and routing changes | Config drift |
| WAF changes during an attack | WAF rules |
| Secrets in git history | Remove secrets from git history |
| Logs and packet captures | Anonymize pcap files and logs |
| Who builds security agents | AI SOC and autonomous SOC |
| Security firms and MDR alert data | Security incident data |
| MSSP and incident response case data | SOC case data |
| Traces left by AI agents in incidents | AI SRE incident agent traces |
Common mistakes when a company reviews its incident records
Most problems come from exporting first and sorting later.
- Sending raw logs with no ticket or timeline around them. Logs make sense to a buyer only as part of an ordered incident.
- Deleting a secret in a new commit and leaving it in git history.
- Including client-environment alerts because they sit in the same ticketing system as your own incidents.
- Breaking the join: removing ticket IDs from commit messages, so the fix can no longer be matched to the incident.
- Assuming old logs exist. CISA and its partners recommend that organizations "store their most important logs for at least six months" (CISA), and many keep no more.
What do AI data buyers ask for first?
AI data buyers have asked Origin for the records of a company running and defending its own infrastructure: incidents joined to the change that fixed them (the ticket, the chat thread, the config diff or pull request, and the outcome), configuration and infrastructure-as-code history with commits and reviews going back years, ordered logs or captures of what happened and what was done, and the runbooks, SOC tickets, on-call records and postmortems of the same period. An AI data buyer told Origin in November 2026 that it wants the company's own infrastructure and its own incidents, not a client's unless the client has agreed, and that the company must own the logs and be able to license them.
The first step buyers asked for is small: two or three incidents end to end, plus a slice of the configuration history from the same months. Before that sample is cut, secrets and Terraform state come out, client-environment records are removed or covered by the client's agreement, and personal data is replaced consistently. The incident timeline template shows how to join the pieces. Origin is an independent data partner. Your company decides what is shared and signs any license itself, and a buyer may pay Origin when a deal closes.
Does your company hold records like these? See what your data is worth.
Common questions
Can a company sell its security incident records and config history to AI companies?
Yes, if the records describe infrastructure the company runs itself and the company owns them. Incident tickets, config and infrastructure-as-code history, runbooks and postmortems can be licensed after secrets, personal data and client data are removed. Records from a client's environment need that client's agreement first. Buyers usually start with a sample of two or three incidents.
Is our Terraform history and incident data worth anything to AI companies?
It can be, when the pieces join up. Years of Terraform commits with reviews, incidents from a paging tool and postmortems that explain each fix show how a real system broke and was repaired. Terraform state files and secrets must come out first, because HashiCorp warns state can hold passwords and API tokens.
Which companies have infrastructure incident data that AI buyers want?
Hosting and CDN providers, managed security and SOC providers, network and security appliance makers, SaaS companies with their own security team, and ISPs. Each runs real infrastructure under attack, changes it often and keeps tickets, configuration history and postmortems. Each also has its own line between its records and its customers'.
Can we license incident records from client environments?
Only with the client's agreement or a clear right in the contract. Records produced inside a client's network are usually covered by the services agreement and often by data protection law. Under the GDPR, a processor that decides new purposes for client data is treated as a controller for that processing.
What has to be removed from incident and config data before licensing it?
Keys, passwords, tokens and Terraform state; client and customer data; personal data such as staff names, phone numbers and user IP addresses; evidence under a legal or investigation hold; and details of weaknesses that are still open. Replace values consistently so the incident still reads in order.
What sample of incident data do AI buyers ask for first?
Two or three incidents end to end, plus a slice of configuration history from the same months. End to end means the alert, the ticket, the chat thread, the config diff or pull request that fixed it, and the outcome, all joined by time and ID and cleaned of secrets and personal data.
Sources and further reading
- NIST SP 800-61 Rev. 3: Incident Response Recommendations and Considerations for Cyber Risk Management
- GDPR Article 4: definitions (controller and processor)
- GDPR Article 28: processor
- HashiCorp Terraform docs: Manage sensitive data in your configuration
- PagerDuty docs: Post-Incident Reviews
- CISA alert AA22-131A: Protecting Against Cyber Threats to Managed Service Providers and their Customers
