91视频

Nearly three years ago, which is light years in the world of artificial intelligence (AI), I drafted an article published by the Daily Journal of Commerce on whether AI could be trusted to draft a construction contract. At the time, the blunt answer was 鈥渘o.鈥� I recently revisited this topic using a leading AI chatbot. Despite significant improvements in AI over the past couple of years, project owners, contractors and design professionals should still avoid putting their complete trust in AI to prepare their next construction contract 鈥� at least not without human intelligence reviewing the work of the AI chatbot.

While AI is developing at breakneck speeds, discretion is (still) the better part of valor when it comes to drafting and reviewing construction contracts. In 2023, for purposes of the above-referenced article, I experimented with ChatGPT to see if I could use it to generate an accurate mechanic鈥檚 lien waiver form to pair with a construction contract for a project in California. At the time, it produced a form that looked pretty good to the untrained eye but was not compliant with California law.

When I made this same query to a leading AI chatbot this week, the chatbot produced an accurate form along with a disclaimer to the user accurately stating California requirements when preparing lien waiver forms. When I experimented further and asked the same AI chatbot some basic questions for drafting a change order provision, things did not go as well. Unprompted, the chatbot presented me with a mini-dissertation on change order options and language to seamlessly incorporate into various industry forms, such as American Institute for Architects (AIA) forms. I hadn鈥檛 asked for these suggestions and, while they looked helpful at first blush, it set off an alarm due to the prevalence of 鈥淎I sycophancy鈥� 鈥� a documented pattern where an AI model pursues human approval over accuracy. The chatbot even quoted so-called 鈥渃ore provisions鈥� from an AIA form, which it referred to as 鈥渆xact text excerpts鈥� but 鈥渒ept below the copyright threshold.鈥� Not recognizing the referenced 鈥渆xact text excerpts,鈥� I typed 鈥淎re you sure?鈥� I received this response:

鈥淵ou鈥檙e right to challenge that 鈥� my earlier description was not reliable, and I conflated concepts that are often handled in the Owner鈥揅ontractor agreement (A101/A102/A133) with what A201 actually says. Thank you for catching it. Below is what I can support from published, citable sources.鈥�

While AI has clearly made significant strides in the past couple of years, this anecdote suggests that project owners, contractors and design professionals should still think twice before relying too heavily on AI in connection with their next contract.

Many attorneys have learned this lesson the hard way as courts across the country have been issuing scathing legal decisions chastising (and often sanctioning) attorneys and law firms for submitting court filings containing false, AI-generated case citations or legal opinions. In one case, just last month, a federal judge in New York issued a detailing the many AI-produced errors in court filings by the attorneys. Similar to the AI sycophancy concerns discussed above, the judge in the New York case pointed out that the court filing at issue was 鈥渘oteworthy for its conspicuously florid prose,鈥� as it 鈥渇eatured an extended quote from Ray Bradbury鈥檚 鈥楩ahrenheit 451鈥� and metaphors comparing legal advocacy to gardening and leaving the indelible 鈥榤ark(s) upon the clay.鈥欌��

Notwithstanding the above concerns, AI is making significant inroads into the construction industry and will continue to do so. AI鈥檚 relevance to the construction industry continues to develop daily, and articles are regularly published regarding the uses and associated risks of AI. For example, the Daily Journal of Commerce recently published an article discussing authorship and data disclosure risks for contractors and design professionals when using AI on construction projects.

AI鈥檚 impact on the construction industry is likely to be profound, if it isn鈥檛 already. Project owners, contractors and design professionals should stay up to speed on AI developments and continue to experiment with ways to boost their productivity. But they should also proceed cautiously 鈥� and not forget to ask, 鈥淎re you sure?鈥�

Mario Nicholas is a Stoel Rives LLP partner and a member of the construction and design group in the firm鈥檚 Portland office. Contact him at 503-294-9510 or聽[email protected].

The opinions, beliefs and viewpoints expressed in the preceding commentary are those of the author and do not necessarily reflect the opinions, beliefs and viewpoints of the Daily Journal of Commerce or its editors. Neither the author nor the 91视频 guarantees the accuracy or completeness of any information published herein.