Last week, the White House announced that it had reached an agreement with several artificial intelligence companies to “manage the risks” posed by the technology. The artists, however, say the agreement “does nothing” to protect the creations.
President Joe Biden’s administration on Friday secured “voluntary commitments” from Amazon, Anthropic, Google, Inflection, Meta, Microsoft and ChatGPT creator OpenAI to “help move toward safe, secure and transparent development of AI technology.”
The agreement calls on companies to agree to ensure that products are safe before launching them and will not pose biosecurity or cybersecurity risks. The companies agreed to address “societal risks”, including prejudice and discrimination. It further calls on companies to build trust with the public by using watermarks and other labels to identify audio and visuals generated using AI technology.
But the deal made no reference to the challenges artists fear in the age of AI, such as job loss and intellectual property theft. Artnet News spoke to several artists and art world stakeholders who have been involved in addressing these challenges.
“This document doesn’t say much,” said Mathew Dryhurst.
Dryhurst and fellow artist Holly Herndon launched tools to help artists in the age of AI. These tools include spawnwhich allows users to set permissions on how their style and likeness can be used, and I was shapedwhich allows artists to see if their work has been used to train popular AI art models.
“It’s disappointing that the Biden administration’s deal with AI companies does nothing to protect creators and doesn’t even recognize the act of mass theft that these generators are built on,” said Molly Crabapple, who published an open letter in May calling on publishers to restrict the use of AI-generated artwork. She predicted that only the most elite illustrators will survive the changes in the industry.
“At a time when SAF-AFTRA and WGA are on strike, in part because of the existential risk these generators pose to all of the creative industries, this silence is glaring,” Crabapple added. “[Biden] obviously does not consider it a priority. Artists like Karla Ortiz worked very hard to educate Congress. There are major lawsuits, not just by her, but by celebrities like Sarah Silverman. The administration simply does not listen.
Silverman, a comedian and stand-up actress, has filed a lawsuit against OpenAI and Meta for claiming the companies violated her copyrights by training their chatbots ChatGPT and LLaMA on her book text. Wet the bed. She was joined in the lawsuit by authors Christopher Golden and Richard Kadrey.
“We represent American creators – including authors, visual artists and programmers – whose work has been misused by AI companies as training data without consent, credit or compensation,” said Matthew Butterick, an attorney representing Silverman, in comments emailed to Artnet News.
Butterick noted that three of the seven AI companies involved in the White House announcement “are already charged in litigation over this matter.”
“Notably, these companies have yet to commit to transparency of datasets, which is critical to making AI fair and ethical for everyone, especially the human creators whose work makes these systems so valuable,” he said.
Ben Zhao, professor of computer science at the University of Chicago, is one of the professors who led a research team that developed Glaze, a technology that allows artists to prevent artificial intelligence platforms from stealing their artistic style by subtly altering brushstrokes and palette.
Zhao said the Biden administration “means well” but is “blind to the real risks” that occur, such as the misappropriation of content without consent, compensation or credit. He said such risks are already devaluing entire human industries, destroying the lives of artists and discouraging future generations of artists, writers and musicians.
“The current discussion only includes stakeholders on one side of the struggle and misses the voices of the people these AI models exploit for profit. Without the voices of artists and other creatives, there can be no well-informed or complete understanding of what is at stake,” Zhao said.
The “voluntary” nature of these commitments renders these commitments “meaningless”, he added, calling the provisions outlined in the agreement “ill-defined objectives” that involve technical problems that lack solutions or may be completely unsolvable.
“Take the example of the “watermark” of AI-generated content. There are no robust solutions for watermarking generative context, whether text or images, known to date. Fragile watermarks are ineffective and, even worse, provide a false sense of security,” Zhao said. “Robust watermarks are incredibly difficult to create, especially in adversarial context such as the proposed application scenarios. How hard will these AI companies work to ‘willingly’ build these difficult systems?”
Zhao said the Biden administration must secure new commitments for transparency APIs that “will assess and report on the accuracy of these mechanisms” as well as “real regulation with well-defined and transparent goals backed by plans for testing, enforcement and, if necessary, sanctions.”
“The assumption that big tech will do the ‘right’ thing despite obvious financial disincentives is naïve,” he said.
Crabapple has proposed that the Federal Trade Commission impose “algorithmic bleeding,” a technical term for the destruction of an algorithm, on any company that has trained its models on copyrighted works.
She said the FTC has imposed disgorgement on several companies in the past, including Cambridge Analytica and Weight Watchers, when it was discovered that their algorithms were trained on illegally obtained data.
She also called on the FTC to require companies to only train models on consensually obtained work in the future and “to impose penalties for companies stealing copyrighted works in the future.”
European Union laws on text and data mining and a prospective AI law are “much clearer” on providing guidance on issues such as data collection and model transparency, Dryhurst said. He called the possibility of opting out of training models the “most plausible request” artists could make.
“If a common-sense opt-out isn’t standardized, I’m afraid we’ll start to see a growing balkanization of the open web as artists and businesses choose to build walls and moats around their data,” he said. “We’ve already seen glimpses of this with Reddit and Twitter and I anticipate it will get worse.”
Dryhurst said existing and proposed EU laws are “flawed” but provide “the most practical guidance on how to address artists’ concerns” currently in place.
“Recent meetings of the Senate Judiciary Committee have shown promise in recognizing the value of the ability for creatives to opt out of AI training if they choose,” Dryhurst said.
“I am personally skeptical that the US will take a tougher policy stance than the EU on AI training, given the stakes involved. It is crucial to understand that there is currently a race between countries to provide the most hospitable conditions for AI companies, which are expected to play an important role in the evolution of the economy.
More trending stories:
Scientists have developed the whitest white paint ever made, so reflective it can cool surfaces
Follow Artnet News on Facebook:
Want to stay one step ahead of the art world? Subscribe to our newsletter to receive breaking news, revealing interviews and incisive reviews that move the conversation forward.
