AI’s summer of discontent

Ever since the New York Times sued OpenAI in late December 2023 over copyright infringement claims the future of its chatbot Chat-GPT has been uncertain. 

The events of this past summer have brought more bad news for companies developing generative AI models and could bring changes to the internet that could soon prevent chatbots from accessing websites or require LLM chatbots like OpenAI’s Chat-GPT to pay for using copyrighted works. For publishers and media companies, however, these changes could provide a lifeline that helps them survive the era of AI-generated search results. On July 1, the internet infrastructure company Cloudflare announced that it would be introducing features that would that allow publishers to charge AI crawlers for permission to collect copyrighted material from its website. Cloudflare gave its paying and free customers more options to restrict how AI webcrawlers could scrape data from their websites, including only disallowing AI crawlers on pages with ads. In its blog post announcing the changes, Cloudflare also said that it would be “changing the default to block AI crawlers unless they pay creators for their content.” In late August Cloudflare expanded its pay-per-crawl feature to users on a paid plan.

On July 23, President Trump signed an executive order to prevent the government from using “woke AI.” Its stated aim is to require AIs used by the executive branch to be ideologically neutral. The preamble to the order makes it clear what ideological biases this order seeks to remove: “One of the most pervasive and destructive of these ideologies is so-called ‘diversity, equity, and inclusion’ (DEI).” The order also seeks to exclude concepts from academic theories of gender and race including intersectionality and systemic racism. The order claims that these ideas “displace the commitment to truth in favor of preferred outcomes.”

Amy Winecoff and Chinmay Deshpande wrote in an article for The Center for Democracy and Technology that it’s not clear what a “neutral AI” would look like and that there really isn’t a good way to measure the ideology of an AI. They write that the order “paradoxically positions one ideological perspective as the default standard for neutrality, while labeling others as biased.” In early August, lawyers for OpenAI and the Times scheduled a meeting to decide the amount of chats that would be made avilable to the Times for its copyright lawsuit against the AI company. The Times asked for 120 million chats, while OpenAI only wanted to give 20 million.

On August 26, the parents of Adam Raine (who died by suicide on April 11) filed a wrongful death lawsuit against OpenAI. The lawsuit alleges that OpenAI rushed safety testing of its AI chatbot Chat GPT-4o to beat Google’s chatbot Gemini to market. It argues that OpenAI “deliberately bypass[ed] critical safety protocols” and put “user engagement over user safety.”

Since the lawsuit was filed, OpenAI has announced new parental control features to alert a parent if the chatbot detects their child is in “acute distress” (potentially notifying authorities in cases of imminent harm). OpenAI also wrote in a blog post that it would be building a system to estimate the age of a user to prevent underage users from getting responses that are sexually explicit or that describe suicide.

On September 10, a blog post on the website rslstandard.org ( RSL stands for Really Simply Licensing) announced that its new protocol (based on the RSS protocol) would embed a license in HTML that allows creators to set the terms under which AI companies could use their content. Formed alongside the new protocol is a new collective rights management organization (RSL Collective), which follows in the footsteps of similar organizations that collect licensing fees for digital music. The RSL Standard says that it has options for AI companies to access content, including paying each time they scrape a site, paying each time the content is referenced, or paying a recurring fee for access. The RSL License also allows the terms of creative commons licenses to be honored by requiring attribution for content and specifying which types of organizations can use that content.