Reviewed by Jonathan West · Updated Jul 4, 2026

Sakana Marlin for Contract Review

An honest answer: Marlin is not a redlining tool, but it can research the context around a contract.

Reviewed by Jonathan West · Updated Jul 4, 2026

Sakana Marlin is an autonomous research agent from Sakana AI in Tokyo. You give it a research topic in natural language, it runs on its own for up to roughly eight hours, and it returns a strategic report with an executive summary, strategic options, primary-source citations, and slides. It is not a chatbot and not a language model you converse with.

If you came here looking for a tool that reads a contract and marks up clauses, this is the honest part: that is not what Marlin does. Marlin is a deep-research deliverable tool, not a clause-by-clause review or redlining engine. Pretending otherwise would set you up to misuse it.

This page explains what Marlin genuinely can do around contracts — research market-standard terms, the regulatory context of a contract type, and data-room due-diligence background — and where to look instead for actual clause review and redlining. It also covers the confidentiality and human-review duties that apply.


Marlin does not redline contracts

Contract review, in the sense most lawyers mean it, is clause-by-clause: reading a specific agreement, flagging risky or non-standard terms, and proposing edits. That is an interactive, document-in-hand task. Marlin is built for the opposite shape of work — a topic goes in, hours pass, and a research report comes out.

Marlin does not open your draft agreement and mark it up. It does not propose redlines, track changes, or negotiate clauses. Using it as if it did would waste an hours-long run and produce the wrong kind of deliverable.

Marlin is not a redlining tool. If your task is 'read this contract and mark up the clauses,' Marlin is the wrong tool by design.

Choosing between a research agent and a contract-review tool? Layer3 Labs helps regulated firms match the tool to the task.

Book a Consultation

Where Marlin does fit around contracts

There is real, honest value in the research that surrounds a contract. Before or alongside review, you often need to know what market-standard terms look like for a given deal type, how a regulatory regime constrains a particular contract, or what the background looks like across a data room in a transaction.

Those are research questions, and they suit Marlin. It can survey how a class of agreements is typically structured, map the regulatory context around a contract type, and produce background research for data-room due diligence as part of a transaction.

  • Researching market-standard terms for a deal or contract type.
  • Mapping the regulatory context around a specific kind of agreement.
  • Data-room due-diligence background research in a transaction.

What to use for actual clause review

For the redlining and clause-level review Marlin does not do, use a different tool class. A chat or drafting model can help you interactively read and mark up a specific clause. A dedicated contract-review tool is purpose-built for clause extraction, playbook comparison, and redlining at scale.

A sensible pattern is to let Marlin handle the upstream research — what is market-standard, what the regulatory context is — and then take that context into a drafting model or contract-review tool for the actual document work.


Confidentiality and privilege

Contracts and data rooms are among the most sensitive material a firm handles. Before putting any of it near a third-party AI service, apply ABA Model Rule 1.6. Frame research questions generically where you can — market-standard terms and regulatory context rarely require the client's actual agreement.

For data-room due diligence, be especially careful: transaction materials are typically confidential and often subject to an NDA. Review the vendor's data-handling terms, confirm what happens to inputs and outputs, and obtain any needed consent before disclosing deal-specific documents.


Limitations and mandatory human review

Even for the research it is suited to, Marlin can produce confident output that is wrong, including citations that do not exist or misstate a source. Courts have sanctioned lawyers for filing AI-hallucinated cases, as in Mata v. Avianca. Every citation and factual claim in a Marlin report must be verified by a lawyer against the primary source.

Marlin is not a lawyer and gives no legal advice. Its research on market terms or regulatory context is a starting point, not a conclusion. Under ABA Model Rules 5.1 and 5.3, the supervising lawyer remains responsible for verifying the output and for the substantive judgment about any contract.

Whatever Marlin returns about market terms or regulation, a lawyer must verify the sources and make the substantive call. It informs review; it does not perform it.

Frequently Asked Questions

  • No. Marlin is an autonomous research agent, not a clause-by-clause review or redlining tool. It does not open your agreement and mark it up. For redlining, use a drafting model or a dedicated contract-review tool.
  • It can research the context: market-standard terms for a deal type, the regulatory environment around a contract, and background research for data-room due diligence in a transaction. That research then feeds the actual review, which happens elsewhere.
  • A chat or drafting model for interactive, clause-level help, or a dedicated contract-review tool built for clause extraction, playbook comparison, and redlining at scale.
  • Only with care. Transaction materials are typically confidential and often under NDA. ABA Model Rule 1.6 applies. Review the vendor's data-handling terms, frame questions generically where possible, and obtain any needed consent before disclosing deal documents.
  • Yes. Marlin can produce confident but false output, including bad citations. Lawyers have been sanctioned for filing AI-hallucinated cases (Mata v. Avianca). Verify every citation and material claim against the primary source.

Match the right AI tool to contract work

Book a free 30-minute AI workflow audit with Layer3 Labs. We will help you separate contract research from clause review and pick the right tool for each — without exposing confidential deal material.

/ai-workflow-audit