ProbusAI

The practice

You deal with the people who sign the work.

In a market with no approved-auditor list and no licence, the accountability of a named individual is the only thing standing behind a signature. So the people are here, with what they actually do and what they are trained in.

2 practitioners · named auditor of record on every opinion

M.O.

AI Auditor & Assurance Professional

Michael leads the technical side of the practice: adversarial evaluation of AI controls, technical classification analysis for counsel, and the opinions that carry a name at the bottom. He treats an AI system as a socio-technical object rather than a model, on the view that an evaluation which only measures the model has not evaluated the system.

He publishes findings against his own builds, because audit work is confidential and a method is better shown than described. Where a finding is inconvenient, it is still written down.

Areas of practice
  • Technical testing
  • AI red teaming
  • Data management
  • Data privacy
Training
  • AIGP-trained (AI Governance Professional)

Published work

M.E.

AI Auditor & Assurance Professional

Michael works on the governance and evidence side of an engagement. His question on any control is the one an opinion ultimately rests on: what was this meant to do, what record exists that it did it, and do the two match closely enough to sign.

His background is in data management and data quality, which is where most AI assurance work actually lands. A model's behaviour is downstream of its inputs, and an evidence trail that cannot account for the data cannot support a conclusion about the system.

Areas of practice
  • AI governance
  • Audit
  • Assurance
  • Data management
Training
  • Trained in AI auditing and assurance

ProbusAI on LinkedIn

DCWP publishes no approved-auditor list and no licence exists. What NYC Local Law 144 requires is independence, and auditor selection is the employer's own judgement, with real liability attached.