How to Spot Risky Contract Clauses With Claude
Most contract review under deadline pressure turns into pattern matching. A clause looks like a hundred others you've seen, so it gets a quick skim and a mental "that's standard," and the review moves on to the next section. That's exactly how a liability shift or a missing carve-out survives three rounds of negotiation without anyone catching it. An associate reviewing their fifth vendor MSA of the week can read an indemnification clause in the same half-second it takes to recognize the shape of the paragraph, without actually parsing whether this specific version differs from the last four.
Warning
"Standard" is a judgment about clauses you've read carefully somewhere else, not this one. If you haven't actually parsed this specific sentence, you're pattern matching, not reviewing.
The clause that costs you later is rarely the one you read twice.
Why a quick skim misses the expensive clause
The problem with skimming isn't laziness. It's that recognition and comparison feel identical from the inside. Recognizing that an indemnification clause looks familiar takes the same half-second as actually checking it against your standard risk position, so it's easy to do the first and believe you did the second. The two experiences feel the same in the moment, which is exactly what makes the gap so easy to miss until it costs something.
Consider a vendor's draft MSA where the indemnification clause has been quietly broadened, "arising from" changed to "arising from or related to," a phrase that looks nearly identical on a skim but meaningfully expands what your client is on the hook for. A reviewer moving quickly through a familiar-looking paragraph is unlikely to catch a two-word change like that, especially buried in the middle of a clause that otherwise matches the template they've seen fifty times before.
Before: skim the indemnification clause, note that it looks like your usual template, move on.
After: paste the clause into Claude with the actual deal context, ask specifically what shifts risk beyond typical, and get the flagged language back with a redline suggestion in less time than the skim took.
That shift, from recognition to actual comparison, is what catches what a skim misses.
A process for reviewing without missing anything
- 1
Give Claude the clause and the deal context
Paste the actual clause, not a paraphrase, along with who the counterparty is and what kind of deal this is. A liability cap that's reasonable for a $5,000 subscription isn't reasonable for a $500,000 services engagement, and Claude can only judge "reasonable" with that context in hand.
- 2
Ask what shifts risk beyond typical, not just 'is this okay'
A vague "does this look fine" invites a vague answer. Ask specifically which liability, indemnification, or termination language shifts risk onto your side beyond what's typical for a vendor of this size, so the review has something concrete to check against.
- 3
Ask for graduated rewrites, not one fix
Request three versions of the weakest sentence, ranging from conservative to aggressive. That gives you room to negotiate down from the aggressive version instead of anchoring on your only option, which matters the moment the counterparty pushes back on your first ask.
- 4
Check the rewrite against your actual fallback position
Compare the suggested language against what you're genuinely willing to accept if the counterparty pushes back. If the conservative version is still stronger than your real fallback, you know exactly how much room you have before you're on the call.
A review pass like this should leave you able to answer:
Which specific clauses shift risk beyond what's typical for this deal size
What your strongest, middle, and fallback rewrite options actually look like
Whether anything material is missing entirely, not just worded poorly
Here's a prompt that puts this into practice:
Here's the indemnification clause from a vendor's draft MSA for a $150,000 annual services contract. Flag anything that shifts liability onto us beyond what's typical for a vendor of this size, and give me three replacement versions of the worst sentence, ranging from conservative to aggressive.
”When the same clause comes back from the counterparty with edits, it's worth running the redlined version through the same check rather than just scanning for whether it looks "close enough" to what you sent.
Here's the clause as they redlined it back to us. Compare it against the conservative version we originally proposed and tell me specifically what changed and whether any of those changes reopen the risk we tried to close.
”Catching what's missing, not just what's wrong
Missing language is harder to catch than bad language, because there's nothing on the page to react to. A limitation of liability section that was never drafted doesn't announce itself the way a one-sided indemnification clause does. You have to know to go looking for it.
Tip
Ask Claude directly what a contract of this type usually includes that this draft doesn't. A missing limitation of liability section, or a missing assignment clause, won't show up if you're only reviewing what's actually on the page.
This is a draft software licensing agreement. Based on what agreements of this type usually include, what sections or protections are missing entirely, not just weakly worded?
”
Inside Claude Tutorial
Comparing against a standard, not just reading closely, is transferable.
Checking a document against what it should contain, not only what it says, shows up in plenty of contexts beyond contracts. The app has a full lesson on it, with practice that applies to any document review.
When the clause survives negotiation and you still have doubts
By the time a clause has been through three rounds of redlines, it's tempting to treat "we're done arguing about it" as the same thing as "it's fine." Those are different questions, and only one of them actually protects your client. Negotiation fatigue has a way of making the fourth round's compromise feel more reasonable than it would have looked on a fresh read at round one.
Common mistake
Accepting a clause because the redline conversation has gone three rounds and everyone is tired of it. Negotiation fatigue isn't the same as the clause actually being adequate, and it's worth a final, separate check before signature, away from the back-and-forth.
Here's the final negotiated version of the limitation of liability clause after three rounds of redlines. Read it fresh, as if seeing it for the first time, and tell me honestly whether it still protects us adequately.
”If the final read surfaces a real concern this late in the process, it's worth asking specifically how much leverage remains to reopen it, since the answer changes depending on how close the deal is to signature and how much goodwill is left on the table.
Your fresh read flagged a real concern with this clause, but we're close to signature and the relationship matters. What's the lowest-friction way to reopen just this one point without reopening the whole negotiation?
”Continue reading
- Writing a Legal Memo That Survives a Partner's First Read: what to do once a clause review surfaces a real question that needs a documented answer.
- Drafting Client Letters That Sound Like You, Not a Template: the next conversation, telling the client plainly what the review found.
- Getting Started with Claude: the underlying habits, context, iteration, specificity, that this entire process is built on.