Tools of the Trade (Secrets): A Trade Secret Playbook for the Post-Non-Compete World?

In our last post, we covered why building a trade secret management program matters more than blanket confidentiality designations. This post is about what a trade secret program makes possible when someone walks out the door, and why companies without one may be left with employment agreements that cannot do that job alone.

Why It Matters

• In jurisdictions that generally prohibit non-compete agreements, including California, such agreements cannot protect the company’s highly confidential information from getting into a competitor’s hands, because even if signed, they will not be enforced by courts.

• Companies therefore need an alternative plan to protect themselves.

• Trade secret law provides the proper framework, but it requires planning and investment to establish protocols for engineering, sales, and legal teams to implement. 
• Litigation success depends on the company’s ability to identify the information at issue as secret, prove that it was stolen, and show that the company took reasonable protection measures before the theft. It is too late to try to create these protections after the theft, and so companies need foresight and strategic thinking long before the theft occurs, or they may be left with no remedy.

Why No Non-Competes?

In 2024, California enacted legislation making non-compete provisions generally void. That was strengthened in 2025, when the legislature closed a related gap by voiding most stay-or-pay terms in contracts. Federal law has not changed the landscape and the FTC’s proposed nationwide non-compete rule was set aside in August 2024. The Defend Trade Secrets Act (“DTSA”) does not override California’s non-compete law, in fact, it expressly does not allow the DTSA to conflict with state law.  The DTSA, specifically 18 U.S.C. § 1836(b)(3)(A)(i)(I), provides that an injunction cannot “prevent a person from entering into an employment relationship,” and any conditions placed on employment must be grounded in “evidence of threatened misappropriation and not merely on the information the person knows.” The next section goes further stating that the order cannot conflict with state laws restricting restraints on trade, which includes California’s non-compete. 

Thus, companies require employees to sign NDAs and provide confidentiality reminders when those employees depart. If a senior engineer takes technical documents concerning unreleased products to a competitor, those agreements may not by themselves prevent access to or use of the information or establish what occurred. Employment agreements do not eliminate the risks associated with employee mobility. However, a trade secret management program can help mitigate these risks by providing additional controls. But, to be effective, such a program must be established and maintained before a dispute arises.

Apple’s recent complaint against OpenAI demonstrates an example of such a strategy in California. It describes controls and records concerning the information at issue. The complaint, filed July 10, 2026, seeks relief under the DTSA without requesting an order barring Chang Liu or Tang Yew Tan from working at OpenAI. The requested relief is limited to prohibiting the acquisition, use, or disclosure of trade secrets and requiring the return of confidential materials. That is, the relief sought is a restriction on conduct involving specified information rather than a larger prohibition on employment, which would be difficult, if not impossible, to enforce in California. 

Trade Secret Injunctive Relief

An injunction directed to preventing or compensating for misappropriation of specified work or technology, however, provides a remedy without formally prohibiting employment in the manner provided by a non-compete clause.

In Waymo LLC v. Uber Technologies, Inc., No. 3:17-cv-00939-WHA (N.D. Cal. 2017), Waymo alleged that Anthony Levandowski downloaded thousands of files relating to proprietary LiDAR technology before leaving Waymo in January 2016. He then founded Otto, a self-driving truck startup, which Uber acquired for $680 million that August; Uber placed Levandowski in charge of its autonomous vehicle program. Waymo asserted claims under the DTSA and California Uniform Trade Secrets Act. In May 2017, Judge Alsup entered a preliminary injunction restricting Levandowski from LiDAR-related work. ECF No. 433. Waymo’s ability to make that showing rested on records its own controls generated. The design file repository was password protected and reachable only through specialized client software, and Waymo encrypted and authenticated all ingress and egress traffic against a regularly audited list of authorized users. Those controls produced a timeline. Intranet search records showed that prior to joining Uber, Levandowski looked up repository login instructions on December 3, 2015. Download records showed that he installed the required client software on December 11 and pulled more than 14,000 files totaling 9.7 GB, including 2 GB from LiDAR subdirectories. A forensic security engineer identified the downloads in October 2016, months before the complaint was filed. The court characterized the resulting record as compelling evidence that Levandowski had taken the files immediately before leaving. The court characterized the resulting record as compelling evidence that Levandowski had taken the files immediately before leaving. Uber later terminated him, and the parties settled during trial in February 2018. The settlement provided Waymo with Uber equity and included a monitored non-use injunction covering Uber’s autonomous vehicle program with independent compliance review.

The preliminary injunction did not prohibit Levandowski from working at Uber; it restricted his work on Uber’s LiDAR-related activities. The court stated: “This order mainly prohibits Levandowski from working on Uber’s LiDAR, a measure Uber has very recently implemented of its own initiative, so the hardship on defendants will be minimal. On the other hand, this will provide considerable protection to Waymo against Levandowski’s potential misuse of its proprietary information in competing technology. This order therefore concludes the balance of hardships tips sharply in Waymo’s favor, at least as to the relief described herein.” Waymo LLC v. Uber Techs., Inc., No. 3:17-cv-00939-WHA, ECF No. 433, at 22 (N.D. Cal. May 11, 2017). The order illustrates the distinction between a restriction directed to specified work and a categorical prohibition on employment.

Apple’s complaint against OpenAI alleges that Liu downloaded proprietary files, continued accessing Apple systems after his departure, and pulled files comprising more than a thousand pages of technical documents while working at OpenAI. The complaint also alleges that Tan, a vice president who had worked at Apple for 24 years and led product design for the iPhone and Apple Watch, co-founded io Products, which OpenAI acquired for approximately $6.5 billion, and now serves as OpenAI’s Chief Hardware Officer. It further alleges that Tan used Apple’s internal project code names during interviews to obtain proprietary information from candidates and directed them to bring props and parts on which they were working. 

The Work That Happened Before the Lawsuit

The best chance of a preliminary or permanent injunction is via a trade secret lawsuit. But what cannot be overlooked is the timing and forethought required to do so. The Apple complaint describes the trade secret program and records on which Apple relies in support of its allegations, in addition to the employment agreements. The complaint describes, among other measures, need-to-know access controls, code-named projects, segmented repositories, authentication monitoring, access logging, rigorous departure procedures, and supplier confidentiality protocols. It alleges that Apple’s logs recorded Liu’s exploitation of a bug and that Apple had a transcript of his message to a colleague. That sort of evidence requires upfront investment in infrastructure and procedure that companies might be hesitant to prioritize and fund. But the complaint illustrates how a trade secret program can generate the kind of contemporaneous evidence that a plaintiff needs to plead its claims. Establishing and maintaining such controls requires investment before a dispute arises.

Trade secret protection and enforcement can be supported by a program that (a) identifies the information alleged to be a trade secret and the basis for protecting it, (b) maintains infrastructure and records to both protect from and monitor misappropriation, (c) supports the company’s ability to assert claims against the departing employee, and (d) supports a request for preliminary or permanent injunction directed to specified projects or information rather than employment generally. 

That record is now being tested. Apple moved for a preliminary injunction in August 2026, seeking to bar the individual defendants and OpenAI from accessing, using, or disclosing its confidential information, and separately moved to expedite discovery. OpenAI has since opposed, and their opposition puts Apple’s own controls at issue. They contend that Apple has not identified any protectable trade secret with sufficient particularity, and that Apple did not take reasonable measures to keep the asserted information secret.

On the second point OpenAI argues that Apple describes its general security program rather than the measures applied to the specific information it asserts, that Apple encouraged employees to use personal iCloud accounts for work files and relied on employees to move those files manually into a designated folder, that Apple escorted employees departing for competitors out of the building the same day they gave notice, and that Apple failed to disable a departing engineer’s access to a document repository and did not check whether it had done so. What the complaint characterizes as the exploitation of a bug, OpenAI characterizes as a failure to revoke access. Apple, for its part, filed a supplemental brief  reporting that a forensic examination of a MacBook returned by Liu uncovered a specific Apple trade secret file that remained on the device and had been accessed while Liu was employed at OpenAI. The preliminary injunction remains pending with a hearing date of October 1, 2026. Our next post will examine the alleged deficiencies and how a company can structure its identification, access, and offboarding procedures so that its protective measures can withstand similar challenges.  

Build It Before You Need It

Critically, a company cannot wait until an employee leaves with highly sensitive information to start thinking about trade secret protection—and cannot rely on employment agreements. By then, as we said in our first post, it is too late. The companies that can assert their rights invest in identification, segmentation, access controls, and forensic monitoring before anyone leaves. 

Forms of protection: 

• A written policy identifying who designates information as a trade secret and on what basis, supported by a maintained inventory of the categories of information claimed.

• Confidentiality agreements drawn narrowly enough to protect specified information without operating as a de facto restraint on future employment, with those obligations flowed down to suppliers, contractors, and other third parties.

• Onboarding and departure procedures, including written confidentiality acknowledgments, exit interviews with individualized reminders of continuing obligations, certifications that company information has been returned, prompt suspension of system credentials, and preservation of access logs and devices before they are reissued or wiped.

• Segmented repositories reachable only through authenticated channels, so that access follows need rather than seniority, provisioned against a list of authorized users that is audited on a schedule rather than at a departure.

• Internal code names for sensitive programs, and repository contents limited to what the company actually claims as secret, since an over-inclusive archive of low-value material invites the argument that the company did not treat the contents as valuable.

• Logging of authentication, access, downloads, exports, and connections of removable media, retained long enough to cover the interval between a departure and its discovery, which is often measured in months.

• Access to customer relationship management systems scoped to assigned accounts, with approval requirements and a retained record for bulk exports of customer, pricing, or pipeline data.

Subscribe To Viewpoints