← Back to Resources

AI Patent Legal Precedents

Today's law is clear: AI can assist in the patent process, but only humans can be inventors. Courts and agencies worldwide have confirmed this principle. Our platform is designed to operate squarely within these guardrails: an on-device scanner identifies potential inventions, AI helps draft robust provisional filings, and attorneys finalize full patents.

Last updated: June 2026, reflects the USPTO's November 28, 2025 revised AI inventorship guidance and recent Federal Circuit decisions including AGI SureTrack v. Farmers Edge (June 2, 2026). This page is informational only and not legal advice.

Recent developments (June 2026)

Why Legal Precedents Matter

As AI transforms how developers build, the law is evolving to clarify what role AI can play in the patent process. Current precedent establishes three core principles:

  1. AI is a tool, not an inventor: inventorship must be attributed to a human.
  2. AI-assisted inventions can be patented, provided humans make the inventive contribution and maintain control.
  3. Provisional filings are the safest place to start: they emphasize timely disclosure and breadth, while leaving final claims and enforceability to attorneys.

Key Precedents

AI Cannot Be Named an Inventor

Thaler v. Vidal (Fed. Cir. 2022)

The Federal Circuit ruled that an AI system such as DABUS cannot be named as an inventor under the U.S. Patent Act, because the statute requires inventors to be natural persons. This case was the first major test of AI inventorship in the U.S. and set a binding precedent that all AI contributions must ultimately be attributed to a human inventor. Read the opinion (PDF)

Supreme Court Denial (2023)

In 2023, the U.S. Supreme Court declined to review the Thaler case, leaving the Federal Circuit's decision as the controlling law. By refusing certiorari, the Court effectively cemented the position that AI inventorship is not permissible in the U.S. patent system. Coverage via IPWatchdog

UK Supreme Court (2023)

The UK's highest court confirmed that only humans may be inventors under the Patents Act 1977. While sympathetic to the growing role of AI, the Court emphasized that legislative change, not judicial interpretation, would be required to recognize AI as inventors. Read the UKSC decision

European Patent Office (2021)

The EPO rejected applications naming DABUS as the inventor, holding that European law also restricts inventorship to natural persons. This decision, echoed by other jurisdictions such as Australia on appeal, reflects a global consensus that AI cannot legally be listed as the originator of an invention. EPO news release

Takeaway: Across jurisdictions, inventorship is reserved for humans, and attempts to assign inventorship to AI have been consistently rejected.

AI Assistance Is Permissible

USPTO Revised Inventorship Guidance (Nov 28, 2025) — current authority

On November 28, 2025, the USPTO rescinded its February 13, 2024 AI inventorship guidance "in its entirety" and replaced it with revised guidance (90 Fed. Reg., Docket PTO-P-2025-0014). The accompanying February 2024 examples were withdrawn at the same time. The revised guidance is shorter, less prescriptive, and reframes how AI-assisted inventorship is analyzed.

Three substantive changes:

  • No separate AI standard. The USPTO now says there is "no separate or modified legal standard" for AI-assisted inventions. Conception by a natural person is the test for any invention, AI-assisted or not.
  • The Pannu factors test is withdrawn for AI scenarios. Under the rescinded February 2024 framework, examiners ran a claim-by-claim Pannu analysis to test each named human's contribution. The revised guidance limits Pannu to its original purpose — analyzing joint inventorship among multiple natural persons. Where one human uses AI, Pannu does not apply (AI is not a person, so there is no joint inventorship to analyze).
  • "Significant contribution" requirement retained. Humans must still contribute "in some significant manner to conception or reduction to practice." The substance survives; only the procedural route through Pannu is gone.

Federal Register notice (Nov 28, 2025) · USPTO announcement

USPTO SME Guidance Update (July 2024)

In its subject matter eligibility update, the USPTO published AI-specific examples (Examples 47–49) illustrating how AI-related inventions may be patentable. These examples highlight that broad functional claims tied to "using AI" may face §101 challenges, but claims demonstrating a technical improvement, such as efficiency or architecture, can pass muster. This gives applicants a roadmap to responsibly integrate AI into patent filings without risking rejection. Federal Register notice

Takeaway: AI involvement is permitted and increasingly expected in the patent process, but humans must remain the inventive agents and claims must be framed around concrete technical contributions.

Provisionals as a Conservative Entry Point

USPTO on Provisionals

Provisionals secure a filing date, require no claims, and are not examined. They are ideal for capturing human-led, AI-assisted ideas quickly while leaving 12 months for refinement. USPTO Provisional Basics

Takeaway: Provisional filings are legally conservative and highly compatible with AI-assisted drafting.

TJTM Technologies v. Google (Fed. Cir. May 5, 2026)

Fresh Federal Circuit application of Alice/Mayo to a software patent. TJTM Technologies, LLC v. Google LLC, No. 25-1218 (Fed. Cir. May 5, 2026) (nonprecedential). The patent at issue (U.S. Patent No. 8,958,853) claimed a mobile-device "inactive mode" that suppressed incoming notifications and sent automated away messages when the phone paired with a vehicle.

The Federal Circuit affirmed the Northern District of California's §101 dismissal in full. Judge Chen, joined by Judges Dyk and Stark, held that the claim "describes a different mode of operation for a mobile phone without reciting any change to the underlying mobile phone technology itself," concluding that "extra user benefit alone does not amount to a 'technological improvement' under our precedent."

The decision is nonprecedential, but reinforces a recurring §101 failure pattern for AI-assisted and software inventions: claims that describe a useful new mode of operation without specifying how the underlying technology is improved. AI features layered onto existing hardware without architectural change face the same vulnerability.

Reported in IPWatchdog, Federal Circuit Affirms 101 Dismissal for Google in Distracted Driving Patent Suit (May 5, 2026). District court initial dismissal was Feb. 2024 with leave to amend; final dismissal was Sept. 2024 after the first amended complaint.

Guardrails for AI in Patents

Permitted (with care):

  • AI used for drafting or expanding disclosures
  • Human conception and oversight documented
  • Recordkeeping of edits and review
  • Disclosing AI assistance when material

Prohibited/risky:

  • Listing an AI as an inventor
  • Filing without human inventive contribution
  • Submitting raw, unreviewed AI outputs

Global Consensus

  • U.S., UK, EPO, Australia: All reject AI inventorship.
  • China & Japan: Policies indicate that only humans can be inventors, while AI can be used as a drafting aid.
Takeaway: The "human inventor, AI tool" principle is emerging as a global standard.

Practical Implications

  • For developers: AI can help uncover hidden inventions, but human contribution is required.
  • For businesses: Provisional applications are the most efficient, risk-managed entry point.
  • For attorneys: AI provides breadth and efficiency at the provisional stage, while legal professionals bring precision at conversion.

Frequently Asked Questions

Does using AI jeopardize inventorship?

No. As long as a human conceived the inventive concept and remains in control, using AI does not undermine inventorship. U.S. law requires inventors to be natural persons (Thaler v. Vidal, 2022). AI may be used to draft, expand, or refine, but the inventive idea must originate with a human.

The USPTO's revised Inventorship Guidance (Nov 28, 2025, 90 Fed. Reg., Docket PTO-P-2025-0014) reinforces this and clarifies the framework: there is no separate AI-specific standard. Conception by a natural person is the test, AI-assisted or not. USPTO announcement.

Best practice: Keep notes, commits, or diagrams proving human ideation and edits.

Should AI involvement be disclosed in a patent application?

It depends. Applicants have a duty of candor (37 CFR 1.56). While AI use does not need to be disclosed in every case, if AI's role is material to patentability, it should be noted.

The USPTO suggests disclosure is appropriate where AI played a meaningful role. Transparency helps prevent future challenges and aligns with examiner expectations.

Best practice: Record AI's role and let counsel decide whether disclosure is necessary.

Can AI-drafted provisionals become enforceable patents?

Yes, with human review and attorney involvement. Provisionals secure early rights but are not examined. AI can generate broad disclosures quickly, but lawyers are essential when converting to full patents.

As long as human inventorship is clear, AI-assisted provisionals can form the foundation of strong, enforceable patents. USPTO provisional basics.

What about AI-assisted patents outside the U.S.?

Global precedent aligns with the U.S.: AI cannot be an inventor.

  • UK Supreme Court (2023): only humans may be inventors.
  • EPO (2021): rejected AI-inventor applications.
  • China and Japan: official policies indicate the same stance.

Takeaway: Assume human-only inventorship worldwide. AI remains a permissible tool.

What's the biggest mistake teams make when using AI in patents?

Treating AI outputs as the invention itself. Common missteps include:

  • Filing disclosures without human contribution
  • Failing to document human ideation
  • Using vague "AI magic" claims without technical detail

Best practice: Treat AI as a drafting assistant, not an inventor. Always document human input, avoid overclaiming, and consult attorneys for conversion.

Resources