Artificial intelligence is making it cheaper to invent. That doesn’t mean, however, that it’s getting any less costly today to turn an invention into a business. That distinction is emerging as a new fault line in the debate over intellectual property and competition policy.
As AI systems become capable of generating scientific discoveries, software and creative works with diminishing human involvement, policymakers face a key question. If the cost of producing ideas approaches zero, does the economic justification for protecting them begin to disappear?
Not necessarily, according to Paul Michel, former chief judge of the U.S. Court of Appeals for the Federal Circuit, and Matthew J. Dowd, founder and managing partner at law firm Dowd Scheffel, who argued the opposite during a discussion hosted by Competition Policy International (CPI), a PYMNTS company.
“If you have a small innovative company, and they come up with a great idea, and that idea happens to be in the area where there are two or three dominant firms, that small innovator only has a fighting chance if they have strong intellectual property,” Dowd said.
The counterintuitive implication is that weakening patents could reinforce concentration by making it easier for incumbents to absorb innovations developed elsewhere.
AI Can Invent a Drug, but Someone Still Has to Finance It
An AI system might identify a promising pharmaceutical compound in seconds, but bringing that discovery to market could still require years of clinical trials, regulatory approvals, manufacturing investments and distribution agreements. The cheaper invention becomes, the more important ownership rights may be to determining who captures its commercial value.
“Even if you assume that the incentive isn’t needed at step one, I suggest it is going to be needed at step two, three, four, five,” Michel said, adding that the financial implications extend beyond pharmaceuticals. AI may reduce the cost of generating intellectual property without reducing the capital required to turn it into commercially viable products.
“These investments of money and other resources tend to escalate sharply as you move down the line from original conceptualization to final product,” Michel said.
Dowd pointed to Anthropic’s reported acquisition of a biology laboratory as evidence of AI developers expanding into physical research infrastructure.
But even vertically integrated AI companies may depend on outside partners for testing, manufacturing and commercialization. Intellectual property provides the contractual foundation for those relationships. Without enforceable ownership, investors may become less willing to finance the expensive stages between discovery and market adoption.
“The property rights allow the transactions of the later transactions, the later developments between Anthropic and subsequent companies,” Dowd said.
Weakening Patents Could Strengthen AI Monopolies
The argument becomes more provocative when applied to competition policy. Patents are frequently criticized for granting exclusive rights that restrict competition. Michel and Dowd said those same rights can provide smaller innovators with leverage against dominant technology platforms.
The implication is that intellectual property protections can function as competitive infrastructure, enabling startups to license innovations without surrendering control to companies with superior capital and distribution.
“By weakening patent rights, you’re actually encouraging or maybe facilitating monopolistic tendencies,” Dowd said.
One shouldn’t assume that established technology companies can supply sufficient innovation themselves, Michel said. For regulators, protecting competition does not necessarily require weakening the property rights of competitors.
Copyright’s AI Problem Goes Beyond Training Data
No discussion of AI is complete without the mention of copyright, which presents a related but distinct challenge. Ongoing litigation over copyrighted material used to train AI systems has raised questions about whether technology companies can build competing products using the creative output of publishers, authors and other rights holders.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
Dowd criticized proposals to broaden fair use protections for AI developers, arguing that some approaches risk excusing alleged past infringement, while Michel identified uncertainty as a more fundamental commercial problem.
“Fair use doctrine is already a mess,” Michel said. “It’s so unpredictable, it’s so subjective.”
Dowd also advocated copyright protection for purely AI-generated works, saying reliable ownership rights could encourage further investment. That proposal conflicts with the existing U.S. human-authorship requirement and remains legally contested.
The Biggest AI Antitrust Risk May Be Getting the Market Wrong
The same uncertainty complicates antitrust enforcement. Developing frontier AI systems requires capital investment, potentially making market concentration an economic feature of the industry rather than necessarily evidence of unlawful conduct.
Meanwhile, competition extends beyond national borders.
“It’s not a national market,” Dowd said. “It is a global marketplace.”
“Antitrust law is very important,” Michel said. “I’m not a critic of antitrust law.”
However, Michel warned that regulators risk damaging innovation when enforcement rests on unproven assumptions about rapidly changing markets.
At the same time, intervention based on economic theories that fail to reflect reality can make competition worse. The unresolved question is how policymakers distinguish legitimate economies of scale from anticompetitive behavior without undermining the incentives that support investment.
For AI companies, investors and small innovators, the stakes extend beyond legal doctrine.
AI may democratize the ability to invent. Whether it democratizes the ability to profit from invention will depend partly on who can establish, defend and commercialize the rights to what gets created.
Watch the full CPI TechReg Talks interview with former federal appeals court Chief Judge Paul Michel and Dowd Scheffel’s Founder and Managing Partner Matthew Dowd to hear more about:
- Why AI could make invention cheaper without making innovation less expensive.
- Why weakening patents could make Big Tech even more powerful.
- Why aggressive AI antitrust enforcement could backfire.
For all PYMNTS AI coverage, subscribe to the daily AI newsletter.