Non-compete Enforceability in the AI Sector: Practical Takeaways from Palantir Technologies Inc. v. Jain

By: Renée Kohler, MacKinzie Neal, Matthew Sumner and Jordan Briggs

In February 2026, the United States District Court for the Southern District of New York granted in part and denied in part a motion for preliminary injunction brought by Palantir Technologies Inc. against three former employees, concluding in part that non-compete language contained in their employment agreements was overbroad. This decision offers a useful lens through which to examine the enforceability of restrictive covenants in the rapidly evolving AI industry. Organizations that continue to rely upon template non-compete provisions drafted prior to the emergence of the current AI talent market may find such agreements inadequate upon judicial review.

The individual defendants occupied distinct roles within Palantir’s organizational structure. Radha Jain served as a senior AI product engineer on Palantir’s flagship Artificial Intelligence Platform (AIP) products. Joanna Cohen, an engineer, led significant healthcare deployments, and Hirsh Jain managed Palantir’s healthcare portfolio. In August 2024, Hirsh Jain departed Palantir to found Percepta, and the remaining defendants followed thereafter. According to the complaint, Percepta recruited at least ten former Palantir employees within months of its founding, with former Palantir staff comprising nearly half of its workforce.

All defendants had signed a Proprietary Information and Inventions Agreement with Palantir that included restrictive covenants. Specifically, the restrictive covenants at issue prohibited the defendants from: (1) performing their Palantir job functions at any “entity engaged in the same or similar business” for a period of one year and (2) from soliciting Palantir customers or employees, for two years. The defendants contended that these restrictions would, in practical effect, preclude them from employment at any company utilizing AI to serve its clients. Given the current breadth of the AI industry this characterization presented a significant interpretive project for the court.

In the now-unsealed opinion, the court concluded that the non-compete language was overbroad and accordingly declined to grant preliminary injunctive relief on the non-compete claims asserted against both Radha Jain and Cohen.

The court did not, however, deny Palantir’s motion in its entirety. Injunctive relief was granted on employee non-solicitation claims against both Hirsh Jain and Radha Jain, and on confidentiality claims against Cohen. The latter claims were based on allegations that Cohen transmitted confidential Palantir documents to herself via Slack on the day following her resignation announcement.

The decision implicates several doctrinal and practical considerations relevant to the drafting and enforcement of restrictive covenants in the AI sector. The following observations merit particular attention.

• The perils of overbroad competitive descriptions. The non-compete language used by Palantir was sufficiently expansive to encompass virtually any entity operating in the fields of AI, data analytics or enterprise software. This outcome underscores a recurring tension in restrictive covenant jurisprudence more generally: the broader a company’s commercial footprint, the greater the precision required in delineating the scope of competitive activity subject to restriction. For platform companies similar to Palantir that operate across a number of different verticals, a generalized “similar business” restriction is unlikely to withstand judicial scrutiny. More defensible drafting would identify specific product categories, market verticals or named competitors rather than relying upon broad categorical descriptors. This need for precision is especially critical when a company operates on a global scale.

  
  

• The nexus between restrictions and protectable interests in non-competes. Under New York law, courts evaluate whether a non-compete is reasonably necessary to protect legitimate employer interests, including trade secrets, confidential customer relationships and specialized training. Where a material disconnect exists between the scope of the restriction and the interest it purports to safeguard, the provision becomes vulnerable to legal challenge. A well-drafted non-compete should articulate this connection with specificity: what proprietary information or training does the individual have, and what constitutes the narrowest restriction adequate to protect that interest?

  
  

• The independent enforceability of non-solicitation and confidentiality provisions. The outcome in this case illustrates the importance of structuring customer and employee non-solicitation and confidentiality provisions as independently enforceable obligations rather than treating them as ancillary to a non-compete. Palantir failed to obtain relief on its non-compete claims yet secured meaningful injunctive relief through its non-solicitation and confidentiality clauses. This result suggests that, when constructing a restrictive covenant framework, each provision should be capable of independent enforcement, containing distinct definitions, differentiated scope, and, where jurisdictionally required, separate consideration.

  
  

• The importance of establishing a robust evidentiary record prior to litigation. One of Palantir’s most compelling piece of evidence included communications discussing the targeted recruitment of Palantir employees, confidential documents transferred to personal devices and a documented chronology of departures. This aspect of the decision reinforces the principle that employers that anticipate potential breaches of restrictive covenants should undertake evidence preservation and access documentation efforts prior to initiating litigation, rather than relying upon the discovery process to develop the factual record.  Further, using IT forensics can be an important tool when evaluating potential breach of confidentiality and improper solicitation. 

  
  

• Periodic review of restrictive covenants in a dynamic market. Employers operating in the AI sector should conduct regular assessments of their existing restrictive covenant language to ensure continued enforceability in light of the rapidly shifting competitive landscape.  In general, non-compete agreements are disfavored by courts in many states, and are statutorily limited or prohibited in some states.  Overly broad non-compete agreements risk further scrutiny by courts in many businesses including the rapidly evolving AI industry. 

This decision warrants continued attention from corporate counsel, private practitioners, and scholars, as courts undertake the task of applying traditional restrictive covenant doctrines to an industry in which the competitive landscape and industry-consuming breadth is expanding at a pace that could render many current contractual instruments and standard language vulnerable. Further, the significant variation in the enforceability standards, statutory frameworks and judicial treatment of restrictive covenants across jurisdictions, practitioners should evaluate the applicable law in each relevant jurisdiction when drafting, reviewing, or seeking to enforce such provisions.

Palantir Technologies Inc. v. Jain et al., No. 1:25-cv-08985 (S.D.N.Y. Feb. 18, 2026)