Context:
- Artificial Intelligence has transformed our everyday lives, including in the patent ecosystem, where it is being used for everything from patent drafting to prosecution and monetization. AI is also being used to accelerate discovery, design software, and even shape new chemical compounds, and the emergence of tools such as Deep IP, Patent Bots, and Solve Intelligence has helped patent professionals streamline their work.
- But, as noted in an article on ai fray earlier this year, AI can mislead innovators, such as through hallucinations (fabricated concepts, fake prior art, or unworkable ideas masked by technical jargon), as well as the consequences of such hallucinations (June 20, 2025 ai fray article). And, when engineers and patent professionals become over-dependent on AI, they “risk more than disappointment”, according to Dr.-Ing. Robert Klinski, a patent attorney and founder of Germany-based firm PATENTSHIP.
What’s new: “With great power comes great danger,” insitro’s IP head Feng Zhou said on a panel at the 12th Annual IP Dealmakers Forum U.S. (October 16, 2025 ip fray article) in Austin, Texas, on Wednesday. Mr. Zhou was speaking at an inaugural one-day event on how AI and emerging technologies are transforming IP transactions, strategy, and monetization alongside the IP Dealmakers Forum. While AI tools are knowledgeable and can generate things that seem plausible, they tend to hallucinate, he said, adding: “So really the biggest challenge with AI is to work with that hallucination and reduce those errors.”
Direct impact and wider ramifications: Mr. Zhou was joined by several other IP professionals in voicing his concerns at the IP and AI event, including Bridget Smith, Assistant General Counsel, IP, Relativity Space, who warned attendees about the harm of malicious (and benign) data poisoning. While these issues are problematic, according to the panelists, they can also be managed. And patentees should continue to trust the experience they have over the information that is being fed to them by machines, fellow panelist Kirk Cesari, Senior Counsel of IP Transactions at Hewlett Packard Enterprise, said.
Supercharging IP commercialization
At the 12th Annual IP Dealmakers U.S. in Austin, Texas yesterday, several IP professionals shared how they are using AI to help in their practices, including both in-house and externally, from drafting to prosecuting to enforcing.
The first panel, entitled “AI-Powered Monetization: How You Really Should Be Using AI”, was moderated by Charles Eldering, CEO of CAsE Analysis, and saw Kevin O’Riordan, VP, Associate General Counsel, Corporate, Tennant Company, note that AI has allowed the 150-year-old company to identify patterns where it has taken a good innovation, translated it into a patent, and then realized real revenue based on it.
“We’re very lean,” he said, “so AI has allowed us to supercharge a lot of the really substantive tasks and leverage our intangibles with really good people to figure out where we go from here.”
AI is a “force multiplier”, Mr. O’Riordan added, as it “opens the aperture” for what’s possible and what we may not know.
Co-panelist Tom Hochstatter, Co-Founder & President at Techson IP, added that AI allows companies to scale up in all aspects of IP, while HP’s Mr. Cesari noted that it has been very effective in helping the company build a valuation of its 4,000-strong patent portfolio.
The discussion around the benefits and risks also came up on a separate panel earlier in the day, entitled “AI-Powered Disruption: A Deep Dive Into Use Cases And Success Stories”. The panel was moderated by Natalie Parker, VP of Contingents Risk at CAC Specialty.
Alex Stroe, Co-Founder & CEO of Patent Watch – an AI-powered patent infringement detection and automated claim chart generation platform – said there are many use cases now:
- IP protection
- Selling licenses
- Tech transfer angle (universities and research institutes looking for someone to build this out for them)
- Commercialization
- M&A
But the recurring theme, Mr. Stroe emphasized, is enabling companies, no matter their use case, to cover their full portfolios. IP teams are not scaling with IP portfolios right now. Those with 100 patents may have one IP person, while those with 30,000 might have a team of 10 or 20, but they are never big enough to cover these portfolios. He noted that nobody is losing their jobs in this scenario:
“The human still has to come in, as infringement is still somewhat subjective. But we let companies come in, cover their full portfolios, show where the value is and where they can defend.”
Meanwhile, fellow panelist Zhou, who heads IP at AI-assisted drug development company insitro, asserted that not learning how to use AI for work is “irresponsible” and “lazy” as it “truly improves” the quality and quantity of his work “tremendously”.
“I hear people saying, ‘it’s not ready, we tried it, it can’t do it’,” he noted, “but that’s only true if you don’t learn how to use the tools and how to take advantage of them and combine their strengths to solve your problems.”
Where AI falls short
Mr. Hochstatter told attendees that he believes the greatest shortfall is more psychosocial than technological:
“There is a collective of scientific scepticism. We have this human nature to race to the end and work backwards. I know there is a gap to the penultimate finish line, but it’s providing value right now, and it’s only day one of this multi-decade advance.”
Mr. Cesari, meanwhile, said that the risk with huge companies like HP is that while they can have trainings and put out policies, they are under no illusion that every single one of their 60,000 employees will read those policies. Groups and entire business units will bring in tools that have not gone through the procurement process, but once someone downloads and uses a tool before it has been vetted for confidentiality, that’s when things will go wrong.
And, for Mr. O’Riordan, the largest issue is the level of embellishment, skewing, and bias that AI is capable of. “We’ve noticed that some of our tools will continue to skew into patterns we’ve identified that may not actually be there,” he said.
In the earlier panel, Tony Trippe, Senior Program Leader, AI Initiatives & IP Strategy Manager at Owens Corning, also warned attendees about embellishments. And, while Mr. Zhou had been very supportive of leveraging AI, he emphasized that “with great power comes great danger”.
Hallucination was his greatest concern, but he also discussed the shortfalls of using AI for translation. While it is a much less expensive route (some patent translations cost upward of $25,000 each, according to Mr. Zhou), it often commits technical glitches – such as skipping an entire paragraph or sentence.
So where do we go from here?
“There is no level of perfection that can be attained,” Mr. O’Riordan emphasized when asked how companies can steer clear of these pitfalls. But taking accountability and tempering one’s excitement are both very important, he added:
“In the beginning, we started running into the dark as fast as we possibly could – and while it’s important to stay ahead, it’s also good to slow down occasionally and look around: see where the gaps and holes are.”
According to Ms. Smith, the best way to use AI is to treat it like a child and guide it down the path. This means inserting your expertise into the workflow. Because the more you outsource and the larger your queries for AI, the more it can keep leaning into something that gets you farther from the answer that you want, Ms. Smith said.