Industries

AI drafts, discovers, and now effectively testifies. Courts are asking whether it can be trusted.

Legal AI moved from novelty to infrastructure in three years. Research, drafting, discovery review, due diligence, and expert analysis all run through models now, and the profession learned early what unverified output costs. Sanctions for fabricated citations arrived in 2023, and the rules of evidence are catching up.

Proposed FRE 707Sanctions exposureE-discoveryProfessional duty
Assess your legal AI →
Where AI acts in the matter

The decision point in professional work

The AI acts when it surfaces the controlling case, marks a document responsive or privileged, drafts the clause a client signs, or produces the analysis an expert relies on. In each of those moments, a professional's duty attaches to output the professional did not produce and often cannot fully inspect.

A court does not ask whether the tool was impressive. It asks whether the output was reliable, and who verified that before it was relied on.

What keeps you exposed

What keeps firms and professionals exposed

Fabricated authority

The citation that does not exist

In 2023 a federal court sanctioned lawyers who filed a brief containing cases invented by a chatbot, and courts have repeatedly sanctioned similar filings since. Verification duty sits with the professional.

Machine-generated evidence

A rule is coming for AI output

Proposed Federal Rule of Evidence 707 would subject machine-generated evidence to reliability scrutiny of the kind applied to expert testimony. Reliability will need a foundation, not a vendor claim.

Privilege and confidentiality

The model saw the client's documents

Discovery and drafting tools process privileged material at scale. How the system behaves with that material is part of the duty of confidentiality.

This is already happening
Proposed Federal Rule of Evidence 707 would require machine-generated evidence to meet reliability standards comparable to those for expert witnesses. Documented, independent evaluation of how the system behaves is exactly the foundation that scrutiny will demand.
Proposed FRE 707 · Judicial Conference Advisory Committee
How AVAAS adds value

Evidence built for the courtroom and the engagement letter

Is the output reliable enough to rely on?

Behavioral evaluation measures fabrication, omission, and consistency on matter-like workloads, producing the reliability record a court or client can examine.

Does it hold privilege and confidentiality?

Scenario testing probes how the system handles privileged and confidential material, including what it retains, repeats, and reveals.

Can the firm show diligence?

Certification produces documented, third-party evidence of conformity to a published standard at the decision point, dated before the matter it protects.

Law firms, ALSPs, accounting and consulting firms, and legal technology vendors all now stake professional duty on model behavior. An independent grade turns that bet into a documented judgment.

Related AVAAS coverage: Evidence Reliability Screener · Justice & defense · Background & reporting.

Build the reliability record before a court asks for it.

Tell us where AI enters your matters and workflows, and we will scope an evaluation that stands up to the scrutiny your profession is built on.

Ready to start now? Certify Your AI →  or  email [email protected]