Harvey changelog

Harvey launched Legal Work Orchestration, which lays out a matter as phases, tasks, assignees and review gates inside a Space. Agents run defined tasks with their scope, inputs and reviewer set in advance, and a named lawyer approves each output before the next step starts. Tasks and notifications are rolling out across workspaces now, status can be updated by replying to an email, and reusable matter templates are in early access.

Buyer relevance. It puts a defined human approval point on every agent task and a visible record of who reviewed what. That is the question risk and ethics committees ask before letting agents run multistep matter work.

Harvey released a Motion to Dismiss workflow built with LexisNexis. A planning agent reads the uploaded complaint and proposes up to eight grounds for dismissal rated by strength, the lawyer approves or edits the plan, and parallel agents research each ground in LexisNexis by jurisdiction, producing a memo of three to five cases validated through Shepard's. The draft motion includes tables of contents and authorities and follows the forum, such as Rule 12(b)(6) in federal court, a California demurrer or a New York CPLR 3211 motion.

Buyer relevance. Litigators get a first draft grounded in Shepardized authority rather than model memory, with a planning step where the lawyer decides which grounds go in. Access requires Ask LexisNexis or Lexis+ with Protégé alongside Harvey, and a Motion for Summary Judgment workflow is next.

Harvey's release notes list Agentic Search for DMS covering SharePoint, OneDrive and Google Drive. Harvey can now search a user's connected content in those repositories directly and find the relevant documents without the user first locating and selecting files. It extends the agentic iManage Search released on 2 Sep to Microsoft and Google document stores.

Buyer relevance. Firms that keep matter files in SharePoint or Google Drive rather than iManage or NetDocuments can now let Harvey find source documents itself, which removes manual upload and selection steps and widens what answers can be grounded in.

Harvey says its agent can now search the US Case Law source by publication status, date filed, court, citing references and cited by references, and can review Bluebook citations and isolate specific reporter page breaks. Harvey extracts citing and treatment relationships across the corpus so the agent follows citations between opinions before answering with inline citations. Runtime precedential treatment warnings and full brief cite checking are described as upcoming work, not shipped.

Buyer relevance. Citing and cited by search plus pinpoint page handling move Harvey toward citator style research inside the platform, but buyers should keep a separate good law check until the announced treatment warnings ship.

Harvey now lets users hand any action in a Review Table to the Harvey Agent in plain language. A team can turn a free text column into a fixed classification and rerun it, import metadata from an attached export as new columns, bulk edit cells, and flag and assign rows to a reviewer, all from one chat. Harvey's worked example is a privilege review: reclassifying the privilege column, mapping custodian data from a Relativity export and routing flagged rows to a named reviewer.

Buyer relevance. Review tables are where document review actually happens in Harvey, and until now every change meant rebuilding columns by hand. The privilege example is the one to test. Ask how the agent's edits are logged, and whether rerunning a column keeps the cells a reviewer has already verified.

Harvey introduced Contract Review Agents, in early access through a waitlist. A team builds an agent from its executed contracts, negotiation guidelines and playbooks, then answers a guided interview about how it applies those standards in practice, such as the liability cap it proposes and the fallback it will accept, or which clauses always escalate; the agent then delivers a first pass review of each redline with recommended edits, reasoning and citations to the sources behind each decision. Precedent benchmarking compares a counterparty's proposal clause by clause against positions the team has accepted in executed contracts, and the agent flags where recent negotiated outcomes have drifted from its own standards and proposes an updated position for the team to review.

Buyer relevance. This is the difference between a playbook and a precedent bank made into a product. A playbook records the stated position; the executed contracts record what the team actually accepted, and in house teams have always negotiated from the second while writing down the first. The drift detection is the feature to watch, because an agent that can tell a team its playbook is six months out of date is doing a job that nobody currently owns.

Harvey's 9 September release note puts user level Memory into the product across the Assistant, the Word add in and Playbooks, two weeks after its 26 August early access announcement. A memory is a standing instruction such as a drafting style, a negotiating posture or a jurisdiction of admission, saved on request or when a prompt signals a preference, confirmed on a card at the point of saving, and then applied in future sessions across the web app, Outlook and Word, with Harvey citing the memory it used the way it cites a source. Every memory sits in Personal Settings where it can be edited, searched, deleted or switched off, admins control whether Memory is on and for whom, and Harvey states that individual memories are never used to train models.

Buyer relevance. Memory is where a legal AI product either earns or loses the confidentiality argument, because what it remembers about a lawyer's clients and positions is exactly what a firm does not want quietly accumulating. Harvey has chosen the opposite of the consumer pattern: nothing is stored without a visible confirmation, everything stored is listed and deletable, and the vendor has said in writing that memories do not train models. Scope today is the individual user only; memory shared across a matter or an organization is described as next, and that is the version that will need a policy.

Harvey released Agentic Vault Search to general availability and opened Agentic Vault Organization in Early Access. Search lets a user attach a whole Vault, or work inside a Space with Vaults attached, and have the Harvey Agent locate the relevant documents from signals such as folder placement, naming convention and uploader rather than exact keywords, operating under the user's own Vault permissions and ethical walls with every file it touches logged to the workspace audit trail. Organization takes a plain language instruction such as sorting a data room by counterparty and agreement type, proposes a plan showing every rename, move and tag before anything changes, executes only on confirmation, can undo the most recent set of changes, and never edits file content.

Buyer relevance. The interesting part is the oversight design rather than the search. An agent that restructures a discovery production or a closing data room is exactly the kind of tool a firm's risk committee asks about, and Harvey has answered the three usual questions in the announcement itself: it inherits the user's permissions, it shows its plan first, and it cannot touch content. Organization is early access through a Harvey representative, so only Search is available to buy today.

HarveySep 2, 2026Vendor

Impact

Medium

Verification

Verified

Harvey rebuilt Playbook Review as a multi agent system. The change is architectural rather than user facing: the work of checking a contract against a playbook is now distributed across cooperating agents instead of running as a single pass.

Buyer relevance. Playbook review is the workflow most in house teams buy this category for, so how it is built matters to anyone evaluating depth rather than demo quality. A vendor publishing its own architecture change is also, by itself, a data point on how much it is willing to say about how the product works.

HarveySep 1, 2026Vendor

Impact

High

Verification

Verified

Harvey released Horizon Scanning in early access, a regulatory and legislative monitoring feature. Users can customize which jurisdictions and topics raise an alert, and turn a surfaced change into a downstream artifact such as a memo or a policy update inside the platform.

Buyer relevance. Regulatory monitoring is one of the few legal workflows still routinely done by hand, usually by an associate reading newsletters. Moving it into the same tool that drafts the response is the part that matters here, because the monitoring was never the expensive step.

HarveySep 1, 2026Vendor

Impact

Medium

Verification

Verified

Harvey added opt in access to Anthropic's Claude Fable 5.1, naming the model and its version. The vendor positions it as strongest on complex, document heavy analysis, with the largest gain over Fable 5 in litigation and dispute resolution, and points to public leaderboard results on Artificial Analysis and Vals rather than an internal figure alone. The same post discloses that Fable models may involve data processing practices differing from existing customer agreements, that zero data retention is available to eligible customers only under a time bound exemption pending Anthropic's Enterprise Frontier Safeguards rollout, and that Anthropic offers no regional processing for Fable 5.1, so all data is processed in the United States.

Buyer relevance. Harvey naming the specific model behind a workflow, and letting customers choose it, is the disclosure behavior this axis grades. Most vendors in this index will not tell a buyer which model is running, so an opt in toggle with a named version is worth more than the performance claim attached to it. The retention and residency caveats published alongside it matter as much as the capability: a firm with data residency obligations is being told plainly that this model processes in the United States and that its zero retention position is time bound.