President-elect Donald Trump has surrounded himself with Silicon Valley entrepreneurs — including Elon Musk, Marc Andreessen, and David Sacks — who are now advising him on technology and other issues.
When it comes to AI, this crew of technologists is fairly aligned on the need for rapid development and adoption of AI throughout the U.S. However, there's one AI safety issue this group brings up quite a bit: the threat of AI "censorship" from Big Tech.
Trump's Silicon Valley advisers could make the responses of AI chatbots a new battleground for conservatives to fight their ongoing culture war with tech companies.
AI censorship is a term used to describe how tech companies put their thumb on the scale with their AI chatbots' answers in order to conform to certain politics, or push their own. Others might call it content moderation, which often refers to the same thing but has a very different connotation. Much like social media and search algorithms, getting AI answers right for live news events and controversial subjects is a constantly moving target.
For the last decade, conservatives have repeatedly criticized Big Tech for caving to government pressures and censoring their social media platforms and services. However, some tech executives have begun to moderate their positions in public. For example, ahead of the 2024 election, Meta CEO Mark Zuckerberg apologized to Congress for bending to the Biden administration's pressure to aggressively moderate content related to COVID-19. Shortly after, the Meta CEO said he'd made a "20-year political mistake" by taking too much responsibility for problems that were out of his company's control — and said he wouldn't be making those mistakes again.
But according to Trump's tech advisers, AI chatbots represent an even greater threat to free speech, and potentially a more powerful way to effect control over speech. Instead of skewing a search or feed algorithm toward a desired outcome, such as downranking vaccine disinformation, tech companies can now just give you a single, clear answer that doesn't include it.
In recent months, Musk, Andreessen, and Sacks have spoken out against AI censorship in podcasts, interviews, and social media posts. While we don't know how exactly they're advising Trump, their publicly stated beliefs could reveal the conversations they're having behind closed doors in Washington, D.C., and Mar-a-Lago.
"This is my belief, and what I've been trying to tell people in Washington, which is if you thought social media censorship was bad, [AI] has the potential to be a thousand times worse," said a16z co-founder Marc Andreessen in a recent interview with Joe Rogan. "If you wanted to create the ultimate dystopian world, you’d have a world where everything is controlled by an AI that’s been programmed to lie," said Andreessen in another recent interview with Bari Weiss.
Andreessen also disclosed to Weiss that he has spent roughly half his time with Trump's team since the election happened, offering advice on technology and business.
"[Andreessen] explained the dystopian path we were on with AI," said former PayPal COO and Craft Ventures co-founder, David Sacks, in a recent post on X shortly after he was appointed to be Trump's AI and crypto czar. "But the timeline split, and we're on a different path now."
On All In — the popular podcast Sacks hosts alongside other influential venture capitalists — Trump's new AI adviser has repeatedly criticized Google and OpenAI for, as the show's hosts describe it, forcing AI chatbots to be politically correct.
"One of the concerns about ChatGPT early on was that it was programmed to be woke, and that it wasn’t giving people truthful answers about a lot of things. The censorship was being built into the answers," said Sacks on an episode of All In from November 2023.
Despite Sacks' claims, even Elon Musk admits xAI's chatbot is often more politically correct than he'd like. It's not because Grok was "programmed to be woke," but more likely a reality of training AI on the open internet. That said, Sacks is making it more clear every day that "AI truthfulness" is something he's focused on.
"That's how you get Black George Washington at Google"
The most cited case of AI censorship was when Google Gemini's AI image generator generated multiracial images for queries such as "U.S. founding fathers" and "German soldiers in WWII," which were obviously inaccurate.
An image generated by Twitter user Patrick Ganley, using Gemini.Image Credits:Gemini / Patrick Ganley
But there are other examples of companies influencing specific results. Most recently, users found out that ChatGPT just won't answer questions about certain names, and OpenAI admitted that at least one of those names triggered internal privacy tools. At another point, Google's and Microsoft's AI chatbots refused to say who won the 2020 U.S. election. During the 2024 election, almost every AI system refused to answer questions about election results, except for Perplexity and Grok.
But the Gemini incident stuck out; it caused Google to turn off Gemini's ability to generate images of people — something the free version of Gemini still cannot do. Google referred to that incident as a mistake and apologized for "missing the mark."
Andreessen and Sacks don't see it this way. Both venture capitalists have said that Google didn't miss the mark at all, but rather, hit it a little too obviously. They considered it a pivotal mask-off moment for Google.
“The people running Google AI are smuggling in their preferences and their biases, and those biases are extremely liberal,” said Sacks on an episode of All In from February 2024, responding to the Gemini incident. “Do I think they’re going to get rid of the bias? No, they’re going to make it more subtle. That is what I think is disturbing about it."
"It's 100% intentional; that's how you get Black George Washington at Google," said Andreessen in the recent interview with Weiss, rehashing the Gemini incident. "This goes directly to Elon’s argument, which is that at the core of this, you have to train the AI to lie [i.e., to produce answers like Gemini's]."
As Andreessen mentions, Elon Musk has been outspoken against "woke AI chatbots." Musk originally created his well-funded AI startup, xAI, in 2023 to oppose OpenAI's ChatGPT, which the billionaire said at the time was infected with the "woke mind virus." He ultimately created Grok, an AI chatbot with notably fewer safeguards than other leading chatbots.
"I'm going to start something which you call TruthGPT or a maximum truth-seeking AI that tries to understand the nature of the universe," said Musk in an interview with Fox from 2023.
When Musk launched Grok, Sacks applauded the effort: "Having something like Grok around will — at a minimum — keep OpenAI honest and keep ChatGPT honest," said Trump's AI czar in an All In episode from November 2023.
It's not clear what Trump and other Republicans could do if they actually wanted to investigate OpenAI or Google for AI censorship. It could be investigations by expert agencies, legal challenges, or perhaps just a cultural issue that Trump can press on for the next four years. Regardless of the path forward, Trump's Silicon Valley advisers are not mincing words on this issue today.
“Elon, with the Twitter files, did a privatized version of what now needs to happen broadly," said Andreessen to Weiss, referring to Musk's allegations of censorship at Twitter. "We, the American population, need to find out what’s been happening all this time, specifically about this intertwining of government pressure with censorship … There needs to be consequences."
This article originally appeared on TechCrunch at https://techcrunch.com/2024/12/15/trumps-silicon-valley-advisers-have-ai-censorship-in-their-crosshairs/
