AI contract review: what the tool cannot verify
The two cited sources identify checks rather than assess an AI product. The SRA warning concerns inaccurate AI-generated citations and confidentiality safeguards in England and Wales regulated legal work. NISTA’s PFI guidance describes a UK public-contract review of terms and distinguishes that from assessing actual performance. No named AI review tool was tested for this article.
What a document-based review covers
The NISTA guidance describes a PFI contract review as a structured, document-based assessment of the terms and conditions, rights, obligations, processes and mechanisms set out in the contract. It helps the contracting authority understand what the contract says, how it is structured and how it should be applied in practice, and it gives a clear enough understanding of the provisions to assess compliance, performance or risk where that is necessary.
That PFI definition concerns a public-contract management process, not AI product features. The cited facts do not establish how any specific AI contract-review feature works or what it can verify.
Check citations against their sources
The SRA's first concern is court or other documents containing false or incorrect information, including citations, as a result of the misuse of AI. The notice states that AI tools can produce hallucinations, generating fictitious cases, references or seemingly factual assertions that may appear convincing despite having no basis in fact.
A plausible citation needs checking against its source before it is relied on. The SRA warning supports that verification step; it does not establish that only a person, rather than any technical process, can perform a check.
Assess confidentiality safeguards for the matter
The SRA's second concern is that those it regulates are not fully considering and appropriately mitigating risks to client confidentiality when using AI systems, particularly in relation to clients' personal data or other confidential information. The notice states that both paid-for and free-to-use AI tools may not provide the contractual and technical safeguards needed to maintain client confidentiality.
Price is not the test. The SRA says firms should understand the safeguards that apply and ensure they suit the nature and sensitivity of the information being processed. A firm must make that matter-specific assessment; the notice does not rate any named product.
Reading the terms is not checking performance
The NISTA guidance is explicit that a contract review is not an assessment of how well the contract is being applied or performed. Understanding what a contract requires and establishing what the parties actually do are different exercises.
This is UK public PFI contract-management guidance, not a general statutory rule for every contract and not an assessment of AI. It supports only the distinction between reviewing contractual terms and assessing how well a PFI contract is applied or performed.
| Cited source | What it establishes | What remains to be checked |
|---|---|---|
| SRA warning | AI misuse may generate false citations or assertions | Each citation and assertion in the particular output |
| SRA warning | Paid or free AI tools may lack needed confidentiality safeguards | The safeguards of the product and their fit for the matter |
| NISTA PFI guidance | A PFI terms review is not an assessment of actual performance | Evidence of how that contract is applied or performed |
A verification sequence for reviewers
Read the tool's contractual terms before you upload client material.
Note whether the tool retains, trains on or shares what you submit.
Classify the matter by the sensitivity of the information involved.
Check every citation the tool returns against a primary source.
Confirm every quoted clause against the executed document.
Treat any statement about actual performance as a question for the client, not a finding.
Record the checks you ran and the ones you did not.
What these sources leave open
The cited excerpts do not assess any specific product. The SRA passage addresses AI misuse risks and safeguards in regulated work; the PFI passage describes one public-contract context. This article therefore cannot decide whether a given model can assess enforceability, compliance or client-specific facts.
The supported conclusion is narrower: the cited SRA warning calls for citation and safeguard checks in its regulated setting; the cited PFI guidance separates terms review from performance assessment in public PFI management. The article does not assess what any named AI product can verify.