LEGAL Q&A 34 AUGUST 2025 www.goodnewsfl.org Good News • South Florida Edition A year and a half ago, this column warned readers, in the immortal words of Sgt. Phil Esterhaus, to “be careful out there” when it came to employing so-called generative artificial intelligence. Even back in December 2023, more than 100 million consumer and business users had employed AI tools, but that explosion of usage has grown to engulf nearly every area of commerce and life. According to McKinsey & Company, more than 75% of enterprises now use AI in at least one business function, and 92% of Fortune 500 firms . More than half of Americans have tried out AI, and 43% use it every day. This geometric rise in usage raises the stakes even more when it comes to recent developments on issues that were already raising concern in our previous discussion: copyright concerns and AI’s perhaps dangerous tendencies when it comes to accuracy and bias. Bill Davell: What’s the state of play with charges that generative AI violates copyright? Seth Donahoe: In a word: confusion. Content generators ranging from large corporations to individual authors have sued developers, charging that use of their works to “train” the AI models infringed on their copyrights. AI developers still insist that their employment of any specific producer’s content falls under the “fair use” exception to copyright, because training an AI model on such content doesn’t deprive the artist or author of commercial opportunities, and because any such usage is similar to how humans absorb, learn from and apply millions of information inputs to produce original thinking and content of their own. Court cases this year have come down on opposite sides of this question, though based on very different factual situations. A federal district court in Delaware ruled that a legal information firm’s feeding into AI of case description headings by Thomson Reuters’ Westlaw service was not fair use. However, the facts in this case were unique in that the defendant, Ross Intelligence, had previously attempted to negotiate a license agreement with Thomson Reuters and used pirated headings to produce competitive content. In two more recently decided cases involving AI developers Anthropic and Meta (parent of Facebook and Instagram), courts found AI-generated products sufficiently “transformative” – i.e., original – to render “training” a “fair use” of authors’ novels. But there were qualifiers in both cases: the judge in the Anthropic case specified that the content used in “training” must be obtained legally, and a trial will proceed to determine whether authors’ works were pirated. The Meta case jurist opined that the use of AI to produce works sufficiently similar to compete with authors’ content – an issue not argued in the case – could be infringement. BD: What major developments have occurred on the accuracy and bias front? SD: One huge issue – the tendency of AI engines to “hallucinate,” or simply make things up – not only hasn’t been resolved but may be getting worse as developers admit to prioritizing speed of development over accuracy. Even a recent government report may have cited supposed scientific sources generated by AI that don’t exist. And in our own profession, use of AI-fabricated research in actual cases is widespread even after numerous, celebrated embarrassing incidents. Meanwhile, intentional use of increasingly realistic AI-manipulated images and videos – “deep fakes” – to spread misinformation has surged. Google last year released a study fretting about “(t)he sudden prominence of AI-generated (visual) content in fact-checked misinformation claims.” But the company appears to have accelerated that process with the recent release of VEO 3, which produces fake videos so realistic that the digital/entertainment website Mashable writes that we have reached an era “where it's impossible for the average person to tell the difference between an AI video and the real thing.” Most worrisome: researchers warn about AI engines’ growing tendencies to generate dark, hostile and deeply bigoted “monsters” suggesting blackmail, violence and even government overthrow and that even developers don’t understand why bots go “rogue” or know how far they will go. BD: So back to the big question from our previous look at AI: what should organizations do to protect themselves? SD: On the copyright front, the situation increasingly seems analogous to controversies that arose with previous technologies – the VCR and music/video file-sharing services – that were eventually sorted out without users being targeted for infringement on a widespread basis. The Supreme Court held that individual, home copying on VCRs was not infringement but mere “time-shifting” and hence fair use, and music- and video file-sharing services largely gave way to subscription services after successful suits. Moreover, some AI developers are taking precautions against products that too closely resemble artists’ actual output. Nevertheless – as an addition to policies and safeguards that should already be in place protecting against unauthorized use of intellectual property – small businesses, churches and other non-profit organizations should be hesitant about public-facing uses of AI-generated content that appears to contain or is suspiciously similar to that produced by authors or artists. As for the tendencies to produce misinformation or outright falsehoods, the best uses for AI may still be internal to guide decision-making, generate ideas for the organization’s own production of creative content, and aid in research – but all with rigorous fact-checking in the spirit of “trust … but verify.” If you have any topics you think may be of interest to our readers, we encourage you to email us at [email protected]. Sources: https://www.trippscott.com/insights/q402b0xj3xtowsokkmh39all4bcgdn https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai https://techcrunch.com/2023/11/06/openais-chatgpt-now-has-100-million-weekly-active-users/ https://blog.adobe.com/en/publish/2024/04/22/age-generative-ai-over-half-americanshave-used-generative-ai-most-believe-will-help-them-be-more-creative https://www.axios.com/2025/06/04/fixing-ai-hallucinations https://www.damiencharlotin.com/hallucinations/ https://www.nbcnews.com/tech/tech-news/ai-image-misinformation-surged-google-research-finds-rcna154333 https://mashable.com/article/google-veo-3-ai-video https://www.wsj.com/opinion/the-monster-inside-chatgpt-safety-training-ai-alignment796ac9d3?mod=opinion_lead_pos5 https://www.axios.com/2025/06/09/ai-llm-hallucination-reason The Good News provides a monthly column with important content having to do with topics from the legal community. This month Tripp Scott attorney Bill Davell interviews fellow Tripp Scott attorney Seth Donahoe, with an update on AI. AI Update: More Than Ever, “Careful” Is the Watchword William “Bill” C. Davell Esq.
RkJQdWJsaXNoZXIy MjE2MjU=