Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Applying copyright law to new technology is always a fight. Artificial intelligence is poised to win this round. At the end of 2023, the New York Times sued OpenAI and Microsoft for copyright infringement. The cost of developing generative AI software is already high. If start-ups must also pay for all the online material they scrape, then it could become ruinous. But previous battles between media and tech companies suggest a compromise outcome is more likely.
The NYT complaint claims OpenAI and Microsoft’s “unlawful” use of millions of articles to train large language models threatens the Times’s ability to provide this service. It points to doppelgänger output from OpenAI’s chatbot ChatGPT. The case follows similar complaints from creators including comedian Sarah Silverman. The US Copyright Office has taken comment on possible new rules on AI. Google, Microsoft and Adobe are promising protection for customers.
The NYT case centres on fair use of copyright material. Educational uses, for example, tend to constitute fair use. Creative work has more protection than factual material.
Impact also matters. The NYT could argue that ChatGPT will kill its subscription model. The two are not evenly matched. The Times has about 10mn subscribers while ChatGPT has more than 100mn users. This is not the only mismatch. OpenAI is valued at an estimated 53 times 2023 sales. The Times at just over 3 times.
But there are complicating factors in AI’s favour. Generative AI has been trained on vast data sets. It is not designed to untangle output and verify attribution. OpenAI and Microsoft can also claim that generative AI models do not copy, they learn from material to produce new content.
This is why media companies are really seeking licensing deals, not laws. OpenAI already has a deal with Axel Springer and The Associated Press. Some media outlets are reported to be in talks with Apple. Viacom’s ill-fated attempt to sue YouTube for copyright infringement in 2007 shows why. After losing the case it eventually settled without payment.
YouTube made some concessions. Owners could request content be taken down, for example. A similar agreement might be made between media companies and AI start-ups.
A court case would produce a new way of measuring the value of content creation in the age of AI. But the Times’s case is unlikely to get that far. It may be a notion unfamiliar in tech but this case is more about cash flow than industry disruption.
Read the full article here