Legal AI This Week · Episode 1 · October 16, 2026

The Front Door Wars

Legora and Harvey launch rival products both called Intake for in-house legal teams, Cooley and Google Cloud build a court-filing redaction agent, Steno opens a beta of Litigation Genius, and federal judges in Boston and New York sanction lawyers over unchecked AI citations.

Disclosure: the hosts, Mia Hale and Sam Reyes, are AI-generated voices. The script is written and fact-checked by the AI Tools For Law team.

Audio for this episode is in production. The full script and sources are below.

Stories and sources

News from October 8 to 14, 2026. Vendor claims are attributed to the vendor; we note what each announcement did not disclose.

Legora and Harvey both launch “Intake” for in-house legal teams

Top story

Legora announced Intake on October 13, available now in Slack: its agent asks for missing details, creates a tracked task, and drafts work from the legal team’s own playbooks, with lawyer review on by default. Harvey described its own Intake the same week, spanning email, chat, and business systems, with a live demo set for October 29. Neither announcement disclosed pricing or accuracy figures.

Cooley and Google Cloud develop a redaction agent for court filings

Product watch

Cooley is the launch partner for a Gemini Enterprise agent that recommends redactions of personally identifiable information and other confidential material for attorney review. Attorneys make every redaction decision. The agent is still in development; no availability date, price, or error rate was announced.

Steno opens the beta of Litigation Genius

Product watch

Steno’s new workspace extends its Transcript Genius deposition tool to whole case files, with plain-English questions and page-level citations to source documents. Steno says it is SOC 2 Type II and HIPAA compliant. Existing clients can use the open beta now; no pricing was announced.

Federal judges sanction lawyers over unchecked AI citations

Courtroom watch

In Aguilar v. The Crawford Group (D. Mass.), Judge Angel Kelley revoked a California lawyer’s pro hac vice admission and set fees and costs of up to $10,000 after briefs with fictitious and misquoted authority. In New York, a federal judge who had already imposed case-terminating sanctions ordered an attorney to pay nearly $39,000 to the lawyers who uncovered false AI-generated citations.

Terms explained in this episode

Agent / agentic
AI software allowed to take several steps toward a goal on its own, such as asking a follow-up question, filing a task, and starting a draft.
MCP (Model Context Protocol)
A standard way to connect AI tools to other software, like a universal plug.
PII
Personally identifiable information: anything that points to a specific person, such as a Social Security number or birth date.
SOC 2 Type II
An independent audit of a company’s security controls over a period of months.
Pro hac vice
Permission for an out-of-state lawyer to appear in one specific case.
Hallucination
When an AI model produces confident, plausible text that is false, such as a case that does not exist.
Playbook
A legal team’s written positions on contract terms: what it accepts, what it pushes back on, and what needs escalation.

Transcript

Cold open

Mia: Two of the biggest names in legal AI launched a product with the exact same name, on back-to-back days. That's where we start this week.

Mia: This is Legal AI This Week, from AI Tools For Law. I'm Mia Hale.

Sam: And I'm Sam Reyes. Every week we go through the legal-tech and AI news that actually matters to people who work in and buy for law firms, and we translate the buzzwords on the way.

Mia: One disclosure up front. Sam and I are AI-generated voices. Our scripts are written and fact-checked by the AI Tools For Law team, and sources for every story are in the show notes at aitoolsforlaw dot A I slash podcast. Let's get into it.

Top story: THE INTAKE LAUNCHES

Mia: On Tuesday, October thirteenth, Legora announced a product called Intake. Its press release calls it "an agentic front door for in-house legal teams." That same week, Harvey announced a product for in-house teams. Also called Intake.

Sam: Same name, same week, same customer. The legal press noticed. Legal Cheek's headline was, quote, "Awkward!"

Mia: Let's do what each one actually says it does. Legora's Intake is available now in Slack. An employee posts a legal request in the company's legal channel and mentions Legora. Legora's AI agent reads the request, asks in the thread for anything that's missing, and turns it into a tracked task with an owner and a due date. Then it prepares a first draft of the work using the legal team's own playbooks and precedents.

Sam: Okay, I have to stop you. "Agent." "Agentic." What does that mean here?

Mia: Go ahead, that's your department.

Sam: In plain English, a chatbot answers a question and stops. An agent is AI software that's allowed to take several steps on its own toward a goal. Here, that means asking the follow-up question, filing the task, and starting the draft, without a person clicking through each step. "Agentic" just means "built around agents."

Mia: And the important part, from Legora's release: how far the agent goes is the legal team's decision, request type by request type. By default a lawyer reviews every piece of work before it goes back. For routine requests the team has approved ahead of time, Legora gives the example of a standard N D A, the agent can finish the task and answer the requester itself.

Sam: N D A, non-disclosure agreement. The most routine contract in corporate life. And in-house, by the way, just means lawyers employed by a company, rather than at an outside law firm.

Mia: Legora also says it plans to extend Intake to Microsoft Teams, email, and M C P.

Sam: M C P stands for Model Context Protocol. Think of it as a standard plug that lets AI tools connect to other software, the way U S B lets any keyboard work with any computer.

Mia: Now Harvey. In a blog post dated October thirteenth, Harvey says its Intake extends the legal department's front door across email, chat, and business systems. It names Outlook, Gmail, Slack, Teams, Jira, ServiceNow, and Salesforce. Harvey says it gathers context, then routes and assigns the work by type, risk, and department, and that agents and lawyers send follow-ups and finished work back through the same channels. Harvey says it will show more at a live demo for in-house teams on October twenty-ninth.

Sam: So, for the person signing the contract. What's the actual difference?

Mia: On paper, Legora is live today in Slack only, with other channels planned. Harvey lists more channels, but it's describing a product it's still rolling out, with a demo at the end of the month. Here's what neither one said in its announcement: price, or any measured accuracy for the work the agent prepares.

Sam: And why is everyone fighting over the front door? Legora's release gives one reason. It cites the Association of Corporate Counsel's twenty twenty-six benchmarking report: the median in-house lawyer now supports three hundred sixty-seven employees. Legora calculates that's up twenty-six percent from the year before.

Mia: That's a lot of people asking one lawyer whether they can sign something.

Sam: Whoever owns the place where those requests land gets to see all the work. That's the prize.

Mia: Buyer note. If your company already pays for one of these platforms, ask whether Intake is included or priced separately, and ask which request types you can keep on "lawyer reviews everything." Get the answer in writing.

Product watch

Mia: Product Watch. First, a law firm building its own tool with Big Tech. On October eighth, Cooley announced a collaboration with Google to develop an AI agent on Gemini Enterprise, Google's AI platform for businesses, that reviews court filings for confidential information and prepares proposed redactions.

Sam: Redactions, meaning blacking out the sensitive parts before a document goes on the public docket.

Mia: Right. According to the release, the agent recommends targeted redactions of personally identifiable information, technical information, and other potentially confidential material, for attorney review. Attorneys make every call on what gets redacted before anything is filed. Cooley is the launch partner and is giving product direction from its litigators.

Sam: Personally identifiable information, or P I I, is anything that points to a specific person. Social Security numbers, birth dates, account numbers, home addresses.

Mia: Here's what they didn't say. The agent is still in development. The release gives no availability date, no price, and no error rate.

Sam: And for redaction, the error rate is the whole ballgame. One missed Social Security number in a public filing is a disclosure you can't take back.

Mia: Exactly. Under the federal rules, the duty to redact sits with the filer, not the clerk, so a tool like this changes how fast the work gets done, not who's responsible for it.

Mia: Second, on October thirteenth, the court-reporting company Steno announced the open beta of something called Litigation Genius.

Sam: Open beta meaning it's available to existing customers to try, but it's still officially unfinished, and they're collecting feedback.

Mia: Steno already had a tool called Transcript Genius for analyzing deposition transcripts. A deposition is sworn testimony taken outside of court before trial, and the court reporter's transcript is the official record of it. Litigation Genius widens that to the whole case file: pleadings, medical records, financial documents, discovery responses, images, and transcripts from other court reporters. Users ask questions in plain English, and Steno says answers come with clickable citations to the exact file and page.

Sam: That page-level citation is the feature to test. If the tool says the witness contradicted the medical record, you want to click straight to the page and see it.

Mia: Steno says when one of its court reporting jobs finishes, a case is created automatically with the transcripts and marked exhibits already loaded. It says the product is SOC 2 Type Two and HIPAA compliant, and that it has zero data retention agreements with its AI model providers.

Sam: Three terms, quickly. SOC 2 Type Two is an independent audit of a company's security controls over a period of months, not just on one day. HIPAA is the federal health privacy law, which matters if you're handling medical records in a personal injury case. And zero data retention means the AI provider behind the scenes isn't supposed to store your data or train on it.

Mia: Steno's existing clients can access the beta from their dashboard today. No pricing in the release.

Courtroom watch

Mia: Courtroom Watch. Two sanctions stories this week, and both are about the same thing. Lawyers filing AI-drafted briefs without checking the citations.

Mia: First, Massachusetts. Massachusetts Lawyers Weekly reported on October thirteenth that U S District Judge Angel Kelley in Boston revoked a California lawyer's pro hac vice status and ordered fees and costs of up to ten thousand dollars as sanctions. The case is Aguilar versus The Crawford Group.

Sam: Pro hac vice. Latin, roughly "for this occasion." It's the permission a court gives an out-of-state lawyer to appear in one specific case. Losing it means you're off the case in that court.

Mia: According to the report, the lawyer filed a twenty-page opposition brief within five hours of the other side's motion. Defense counsel moved for sanctions. The court found that about half the citations in that brief had an error, including a case cited six times for a proposition it doesn't stand for, a fictitious citation to Westlaw, the legal research database, and altered quotations. The lawyer acknowledged using A I-assisted legal research.

Sam: And the key line from the judge?

Mia: Judge Kelley wrote, quote, "There is no rule against the use of AI in researching and drafting legal papers, but it must be utilized responsibly. The use of AI does not diminish an attorney's professional and ethical obligations under Rule 11." She also reminded local counsel of the duty to supervise the out-of-state lawyer, though she did not sanction local counsel.

Sam: Rule 11 is the federal rule that says when you sign a filing, you're certifying that the legal arguments are supported by existing law. A made-up case is not existing law.

Mia: Second, New York. Law360 reported on October ninth that a New York federal judge, who had already imposed case-terminating sanctions on an attorney over false A I-generated citations, ordered him to pay nearly thirty-nine thousand dollars to the lawyers who uncovered the problem.

Sam: Case-terminating means the client's case was effectively ended as a sanction. That's about as serious as it gets.

Mia: The pattern is clear. Courts are not banning A I. They're punishing unchecked A I. And the fees are increasingly paid to the other side for the time it took to catch the errors.

Jargon jar

Sam: Jargon Jar. Two terms we used today that show up in almost every vendor demo.

Sam: Hallucination. When an A I model produces something that sounds right and isn't, like a case that doesn't exist or a quote a court never wrote. It's not lying, it's predicting plausible text. Which is why every citation has to be pulled and read by a person.

Sam: Playbook. In contract work, a playbook is a written list of a legal team's positions. What we'll accept, what we push back on, and what needs a senior lawyer. When a vendor says its agent "uses your playbook," it means the AI is following those written rules. So the output is only as good as the playbook you give it.

Mia: Garbage in, garbage out, but with a nicer interface.

Buyer's take

Sam: Buyer's Take. One thing to do this week. If anyone in your firm is using A I to draft anything that gets filed, put a written citation-check step in your process. Every case pulled, every quote matched against the source, initialed by a person. It costs an hour. This week it cost one lawyer his admission in a federal case and another nearly thirty-nine thousand dollars.

Mia: And when a vendor pitches you an agent, ask three questions. What does it cost? What's the measured error rate? And which steps can a lawyer require to be reviewed?

Sam: That's the week. Every story we covered is linked in the show notes at aitoolsforlaw dot A I slash podcast.

Mia: If you want the written version, the newsletter is free at aitoolsforlaw dot A I slash newsletter. Nothing in this show is legal advice, and listening doesn't make you our client.

Sam: Check your citations. We'll see you next week.

Mia: This has been Legal AI This Week.

About the show

Hosts Mia Hale and Sam Reyes are AI-voiced personas; the script is written and fact-checked by the AI Tools For Law team. Not legal advice. Corrections: hello@aitoolsforlaw.ai.

Subscribe with any podcast app using the RSS feed, or get the written version in the newsletter.

All episodes