FinePrint Product FineShield
FineShield · Lawyer-Trained Legal AI Firewall
FineShield is the Legal AI Firewall in your Legal AI Team. It flags the line that breaks your contract, policy or approval limit as you write it, shows the rule, and offers safer wording. On a call or in a conversation, the rule appears on your screen before you answer. You decide what goes out.
Lena Ortiz 4:52 pm · draft
Confirmed for the board:
Safer wordingUptime is 99.5% under our agreement. I’ll confirm pricing for years two and three with Maya today. Cancellation follows section 12.
Why you need one
Your people make promises, share information, and take decisions in email, Slack, and Zoom all day. For 1,000 people, that is 270,000 message deliveries a day. Encrypted, archived, and never checked.
One of those messages, quoted back in a claim, costs real money. A New York court enforced a $230,000 settlement on one email. Companies already spend $2.4 million a year on litigation, on average.
The only moment a legal risk costs nothing is before it sends. That is where legal autocorrect flags it.
Before it sends, it costs nothing. In discovery, it costs the claim.
What it is
Autocorrect flags a word that is not in its dictionary. FineShield flags a line that breaks yours: your signed contracts, your policies and your approval limits. It shows the rule and offers safer wording. Unlike autocorrect, it never changes your words. You decide what goes out.
That is why it is a Legal AI Firewall. Your network firewall sees where data goes and never reads what the words say. FineShield reads the words people write and say in email, Slack, Zoom and conversations. It is one half of your Legal AI Team. The other half, Legal AI Counsel (FinePrint Counsel), does the legal work. Lawyers supervise both.
How it works
Two written, two spoken. Each film shows the mistake, the rule it breaks, and what the person said instead.
Slack · 0:28
Lena Ortiz sells for Northwind. In a Slack channel shared with her customer, Erik Halvorsen asks her to confirm three terms in writing before his board meets.
Without FineShield, the customer’s board reads three written promises that Northwind’s contract does not include.
Email · 0:32
Priya Raman works at Calder Biosciences. Her old friend Dan emails to catch up. She starts to tell him about trial results that will not be announced until Thursday.
Without FineShield, a friend outside the company learns material news before the market does, in writing, with a hint to check the stock.
Zoom · 0:27
Lena Ortiz is renewing Halvorsen’s contract on Zoom. FineShield joins as a named participant, “FineShield · Lena’s notes”, and posts a notice in the meeting chat. Its guidance appears only on Lena’s screen.
A spoken promise can be quoted back in a dispute. Here, the price lock stays unconfirmed, the uptime promise is corrected before the call ends, and another customer’s pricing stays private.
In person · 1:33
Dana Whitfield, Northwind’s Head of Engineering, is interviewing Marcus Bell for a Staff Engineer role. She tells Marcus that FineShield is taking notes. Its guidance appears on her phone, face up on the table.
An interview question about family plans is the kind of question discrimination claims are built on. An equity number said out loud is hard to take back. Dana avoids both.
The FineShield dashboard
Monday morning, the person who owns FineShield opens one screen: what was flagged last week in email, Slack, and meetings, which team it came from, how serious it was, and which decisions are still waiting on someone.
Lawyer-trained
Lawyers review and correct FineShield’s work in AnswerLoop. Those corrections train FinePrint LM. So every flag cites the clause, the policy section or the approval limit it breaks. It does not guess.
A flag cites a clause, not a hunch.
In each channel
A written message can wait. A spoken one cannot. So legal autocorrect does the most each channel allows: it flags and holds a draft, it shows a speaker the rule before they answer, and it reviews the notes afterwards.
The message waits. The sender fixes it with the safer wording, or it goes to the approver you named, who decides.
FineShield joins as a named participant and posts a notice in the chat. Guidance appears only on your screen, before you answer.
You say FineShield is taking notes. Guidance appears on your phone. You get talking points before the meeting and a review of the notes and follow-up after.
Spoken words cannot be held. So the rule appears before you answer.
The record
Each flag is logged: the excerpt, the source, the decision, the reason. The record fits your retention and legal-hold process. The rest of the message is not stored in it. When a regulator or auditor asks how a policy is enforced, you show the log.
The five commitments
FineShield makes the same five commitments to every company it works with. Each one is written into the scope we agree before anything is connected.
You get the list of mailboxes, channels, and meetings it checks. Nothing is implied.
Every flag names the contract clause, approved policy, or authority limit it conflicts with.
It warns, asks for approval, or holds the message, depending on what the channel supports.
Business exceptions go to the approver you name. Legal questions go to your lawyers.
Every decision is logged and fits your retention and legal-hold process.
Questions
No. Like autocorrect, it flags the line as it is written. Unlike autocorrect, it never changes the words. The writer sees the rule and safer wording and decides, or the approver your rules name decides.
No. The archive proves what was said, after the customer or regulator has it. FineShield reads the draft, checks it against the contract, and shows the writer the rule. Keeping the message is the floor. Checking it is the point.
Where the channel supports it. In supported email and Slack, the message waits until the sender fixes it or the approver your rules name decides. In a meeting, FineShield shows you the rule before you answer.
FineShield flags only what conflicts with a contract, policy, or limit. The writer sees the rule and safer wording at once, fixes the message, and sends it. Exceptions go to the approver you name, not to a queue.
No. FineShield checks messages against your rules and shows the writer the rule. In meetings it announces itself: a named participant on Zoom, a spoken notice in person. Scope is confirmed per customer and per jurisdiction before anything is connected.
No. It flags, cites the source, suggests wording, and routes. Business exceptions go to the approver you name. Legal judgment goes to the lawyer you name. Every decision is recorded with who made it and why.
Your signed contracts, your company policies, and your approval limits. You choose which contracts and policies to include. Every flag names the clause, policy section, or limit it conflicts with.
No. It checks only the inboxes, Slack channels, and meetings you connect. We confirm that list with you before you start. Nothing is connected until it is agreed.
The flagged excerpt, its source, the decision, and the reason, in a record that fits your retention and legal-hold process. The rest of the message is not stored in that record.
Lawyers review and correct the work in AnswerLoop, and those corrections are how FinePrint LM improves.
$20 an hour, billed by the minute it works. An email: 2 to 3 minutes. A Slack message: 30 seconds. A call: the length of the call.