AI patent tools cut costs but leave startups dangerously exposed, Landfall IP partners warn

Why AI patent filings often fall short

Michael Drapkin and Nathan Mutter launched Landfall IP in Boulder, Colorado as an AI-native patent law firm. Now, they’re sounding the alarm: startups that rush to file patents using off-the-shelf AI tools may save time and money upfront, but the filings often crumble under scrutiny. Investors, acquirers, or competitors can spot the gaps—sometimes with costly consequences.

Drapkin, the firm’s founding partner, flags three immediate risks. Provisional applications may amount to little more than placeholders. Disclosures made before filing can void patent rights. And tangled ownership records can leave startups unable to prove who invented what—any of which can torpedo a fundraising round or kill an acquisition.

Mutter, who has served as top outside counsel for Oura, Coinbase, and Rubrik, zeroes in on the claims. These define what competitors can be blocked from doing, yet AI tools frequently churn out claims that are either too broad, too narrow, or unsupported by the rest of the filing. The result? Patents that look impressive but offer no real protection.

AI still has a role, Mutter adds. It can turn raw engineering notes into organized invention disclosures, flag prior art, and even suggest variations on a core idea. But deciding whether to patent that idea or keep it as a trade secret—and drafting claims that actually block rivals—remains a job for lawyers.

What investors and buyers really check in a patent

Investors and acquirers don’t count patents—they check whether they’re enforceable, cover what rivals must copy to compete, and belong to the startup without strings attached.

Patent claims must target the minimum features competitors need to build a competing product. Too broad, and they’ll collide with existing patents. Too narrow, and rivals can easily design around them.

The supporting specification—the detailed description of the invention—must leave room for adjustments. Lock it into one narrow implementation, and competitors can work around it. Omit key technical details, and later amendments may get rejected.

Once a startup pitches to investors, demos a prototype at a trade show, or even posts a job listing hinting at the tech, it risks forfeiting patent rights outside the U.S. In most countries, that disclosure voids protection immediately. The U.S. offers a one-year grace period—but only if the filing follows fast.

The hidden costs of cutting corners

Drapkin and Mutter have seen the fallout firsthand. Mutter’s work as top outside counsel for Oura put him in the middle of a patent-infringement lawsuit with Ultrahuman. The dispute hinges on whether Oura’s sensor designs are novel enough to block a rival product—precisely the kind of question AI can’t answer without legal judgment.

The partners point to broader patterns, too. When Coinbase secured its federal trust-bank charter, regulators demanded a full intellectual-property audit. Salesforce’s acquisition of Slack included the same scrutiny. In both deals, legal precision—not the sheer number of filings—carried the day.

Key takeaways

• DIY AI patent filings can leave startups with worthless patents, ownership gaps, or forfeited rights—risks that often surface only during investor due diligence or litigation.
• Deciding what to patent versus keep as a trade secret, and drafting claims that actually block competitors, requires legal judgment AI cannot replace.
• AI can help organize technical documents and flag prior art, but startups still need experienced patent attorneys to avoid costly mistakes.
• Fixing a botched patent filing—through lost exclusivity, failed enforcement, or derailed acquisitions—can cost far more than the legal fees saved upfront.

Related on Lazyfounder

• Oura
• Coinbase
• Salesforce
• Oura sues Ultrahuman again over redesigned Ring Pro, seeks U.S. sales ban
• Coinbase wins U.S. trust bank charter, joining rare group