State-by-State: Where are AI non-competes enforced, and where are they not?

The race for artificial intelligence and machine learning talent is more intense than ever. Meta, Google, Microsoft, OpenAI, Anthropic, AWS, Nvidia, Safe Super Intelligence, and Thinking Machines Lab are all vying for engineers, researchers, and machine learning professionals who can push the boundaries of large language models and generative AI. But whether a non-compete agreement can actually restrict your next move depends heavily on where you live and work. In the United States, the enforceability of non-competes is a patchwork—shaped by state law, local courts, and evolving public policy.

This article breaks down the key differences across major states, highlights how leading AI companies are adapting, and offers practical guidance for both employers and employees navigating this complex landscape.

Key States and Their Approaches

California: The Gold Standard for Talent Mobility

California is famous for its outright ban on non-compete agreements. Under California Business and Professions Code §16600, any contract that restrains someone from engaging in a lawful profession, trade, or business is void. This legal environment has fueled the explosive growth of Silicon Valley and made it a magnet for AI talent. Engineers, researchers, and machine learning specialists at Meta, Google, OpenAI, Nvidia, Safe Super Intelligence, and Thinking Machines Lab can move freely between employers, driving rapid innovation and cross-pollination of ideas.

Massachusetts: Guardrails and Garden Leave

Massachusetts allows non-competes, but only under strict conditions. The Massachusetts Noncompetition Agreement Act (2018) requires that:

• The employee earns above a certain income threshold.

• The agreement is provided in writing and signed by both parties.

• Employers must offer “garden leave”—paying at least 50% of the employee’s highest annualized base salary during the restricted period.

• The duration is limited (usually 12 months), and the scope must be reasonable.

For AI professionals at companies like Microsoft, AWS, or Anthropic, these rules mean non-competes are possible, but only if the employer is willing to pay for the privilege.

Washington: Notice and Narrow Scope

Washington state law requires employers to provide advance notice of non-competes and limits their use to employees earning above a certain salary. The maximum duration is 18 months, and the agreement must be reasonable in scope and geography. For large language model engineers at Meta or OpenAI, this means non-competes are possible, but only if the company follows strict procedural rules.

New York: Reasonableness and Recent Reform

New York courts will enforce non-competes only if they are reasonable in duration, geography, and scope, and necessary to protect legitimate business interests. There is growing legislative momentum to restrict or ban non-competes, especially in tech. For Nvidia, Safe Super Intelligence, and Thinking Machines Lab employees, the enforceability of a non-compete will depend on the specific facts and the court’s view of what is “reasonable.”

Texas: Still Enforceable, But Under Scrutiny

Texas law allows non-competes if they are ancillary to an otherwise enforceable agreement (such as an NDA) and are reasonable in time, geography, and scope. However, courts are increasingly skeptical of broad restrictions—especially in fast-moving fields like AI and machine learning.

How Companies Are Adapting

Meta, Google, Microsoft, OpenAI, Anthropic, AWS, Nvidia, Safe Super Intelligence, and Thinking Machines Lab all operate across multiple states, so they must tailor their employment contracts to local law. In California, they rely on NDAs and trade secret law. In Massachusetts and Washington, they may use non-competes for senior roles, but only with proper notice and compensation. Many are moving away from broad non-competes altogether, focusing instead on protecting specific proprietary information.

Remote work adds another layer of complexity. If an engineer works from California for a company headquartered in New York, California law will likely apply—voiding any non-compete.

Case Studies

• Large Language Model Teams: A Meta engineer in California can move to OpenAI or Anthropic with little legal risk. In Massachusetts, the same move might trigger a paid garden leave period.

• Nvidia Hardware Engineers: In Texas, a narrowly tailored non-compete might be enforceable, but only if it protects specific chip design secrets.

• Safe Super Intelligence and Thinking Machines Lab: These labs often employ hybrid teams across multiple states, so they use a mix of NDAs, non-solicitation clauses, and, where allowed, limited non-competes for key personnel.

Practical Guidance

For Engineers, Researchers, and Machine Learning Professionals:

• Always check which state’s law governs your contract.

• Inquire about the scope, duration, and geographic reach of any non-compete.

• If you’re in California, non-competes are void—don’t be intimidated by unenforceable clauses.

• In other states, ask if garden leave or compensation is required.

For Employers:

• Review and update contracts to comply with local law.

• Focus on protecting trade secrets and customer relationships, not restricting all future employment.

• Consider alternatives like NDAs and non-solicitation clauses, especially for remote or multi-state teams.

Conclusion

The enforceability of non-compete agreements for AI and machine learning talent is highly state-dependent. As Meta, Google, Microsoft, OpenAI, Anthropic, AWS, Nvidia, Safe Super Intelligence, and Thinking Machines Lab compete for the best minds, understanding the legal landscape is essential. For most, the future points toward greater mobility, narrower restrictions, and a focus on true innovation rather than legal barriers.