Methodology

How entries are built

This page exists so the record can be checked rather than trusted. If something here is unclear or wrong, it should be corrected before anything is cited.

What gets logged

An entry records a publicly reported case in which artificial intelligence played a material role in committing, attempting, or enabling a crime. Everything is drawn from open sources. Nothing on this site comes from non-public material of any kind.

Source tiers

Every source on every entry carries a tier. The tier describes where the information originated, not whether it is true.

P Primary
Court filings, indictments, judgments, agency press releases, regulator notices, official statistical publications.
S Secondary
Established news organisations reporting on a primary record or their own reporting.
T Tertiary
Vendor blogs, trade press, aggregators, industry reports. Frequently the only available source, and treated accordingly.

An entry resting only on tertiary sources is not excluded, but it is filterable, so anyone can see how much of the record depends on material with a commercial interest in the subject.

Enforcement status

Most records of this kind stop at the incident. This one records what followed, using whatever the sources show at the time of writing. Status is updated when a case moves; the entry shows when it was last revisited.

  • Reported only — no enforcement action publicly recorded.
  • Investigation opened — an agency has confirmed it is investigating.
  • Charges laid — charges or an indictment filed against a named or unnamed person.
  • Sanctions or seizure — designations, asset seizure, domain or infrastructure takedown.
  • Platform action — account, model, or service-level enforcement by a provider.
  • Civil litigation — private or state civil proceedings underway.
  • Convicted — conviction entered, whether or not sentencing is complete.

Most entries will remain at “reported only” permanently. That is a finding, not a gap in the record.

Figures and statistics

Loss totals, growth rates, and prevalence figures circulating in this field frequently originate with companies selling detection products, and often lack a stated population, observation window, or method. Where an entry repeats such a figure, it states where the figure came from and what it actually measures. Where that cannot be established, the figure is marked unverified and not used.

Naming

No individual is named here unless a court document, indictment, or official agency release named them first. This record does not perform attribution and does not identify anyone not already publicly identified.

What is excluded

  • Anything drawn from non-public sources.
  • Any matter connected to the operator's public-sector employment, and any British Columbia criminal matter.
  • Operational detail, methods, or instructions that would assist in committing an offence.
  • Incident-level narrative in category D, which is covered only at the policy, enforcement, and statistical level.

Corrections

Errors are corrected in place and the change is noted on the entry. Entries are not silently deleted. To report an error, email jason@unio-analytics.com.

Independence

Unio Analytics accepts no funding, sponsorship, or advertising from vendors of detection, identity, or security products, and no payment from any party with an interest in how an incident is characterised. The value of an audited record depends entirely on this, and it is not negotiable.