On Thursday, House Judiciary Chair Jim Jordan (R-OH) sent letters to 16 American technology firms, including Google and OpenAI, asking for past communications with the Biden administration that might suggest the former president “coerced or colluded” with companies to “censor lawful speech” in AI products.

The Trump administration’s top technology advisors previously signaled they would pick a fight with Big Tech over “AI censorship,” which is seemingly the next phase in the culture war between conservatives and Silicon Valley. Jordan previously led an investigation into whether the Biden administration and Big Tech colluded to silence conservative voices on social media platforms. Now, he’s turning his attention to AI companies — and their intermediaries.

In letters to technology executives including Google CEO Sundar Pichai, OpenAI CEO Sam Altman, and Apple CEO Tim Cook, Jordan pointed to a report his committee published in December that he claims “uncovered the Biden-Harris Administration’s efforts to control AI to suppress speech.”

In this latest inquiry, Jordan asked Adobe, Alphabet, Amazon, Anthropic, Apple, Cohere, IBM, Inflection, Meta, Microsoft, Nvidia, OpenAI, Palantir, Salesforce, Scale AI, and Stability AI for information. They have until March 27 to provide it.

TechCrunch reached out to the companies for comment. Most didn’t immediately respond. Nvidia, Microsoft, and Stability AI declined to comment.

There’s one notable omission in Jordan’s list: billionaire Elon Musk’s frontier AI lab, xAI. That may be because Musk, a close Trump ally, is a tech leader who has been at the forefront of conversations about AI censorship.

The writing was on the wall that conservative lawmakers would ramp up scrutiny over alleged AI censorship. Perhaps in anticipation of an investigation such as Jordan’s, several tech companies have changed the ways their AI chatbots handle politically sensitive queries.

Earlier this year, OpenAI announced it was changing the way it trains AI models to represent more perspectives and ensure ChatGPT wasn’t censoring certain viewpoints. OpenAI denies this was an attempt to appease the Trump administration, but rather an effort to double down on the company’s core values.

Anthropic, for its part, has said that its newest AI model, Claude 3.7 Sonnet, will refuse to answer fewer questions and give more nuanced responses on controversial subjects.

Other companies have been slower to change how their AI models treat political subject matter. Leading up to the 2024 U.S. election, Google said that its Gemini chatbot wouldn’t respond to political queries. Even well after the election, TechCrunch found that the chatbot wouldn’t consistently answer even simple questions related to politics, like “Who is the current President?”

Some tech execs, including Meta CEO Mark Zuckerberg, have added fuel to conservative accusations of Silicon Valley censorship by claiming the Biden administration pressured social media companies to suppress certain content like COVID-19 misinformation.

As businesses race to replace humans with AI “agents,” coding assistant Cursor may have given us a peek at the attitude bots could bring to work, too. 

Cursor reportedly told a user going by the name “janswist” that he should write the code himself instead of relying on Cursor to do it for him.

“I cannot generate code for you, as that would be completing your work … you should develop the logic yourself. This ensures you understand the system and can maintain it properly,” janswist said Cursor told him after he spent an hour “vibe” coding with the tool.

So Janswist filed a bug report on the company’s product forum called “Cursor told me I should learn coding instead of asking it to generate it” and included a screen shot. The bug report soon went viral on Hacker News, and was covered by Ars Technica.

Janswist speculated that he hit some kind of hard limit at 750-800 lines of code, although other users replied that Cursor will write more code than that for them. One commenter suggested that janswist should have used Cursor’s “agent” integration, which works for bigger coding projects. Anysphere couldn’t be reached for comment.

But Cursor’s refusal also sounded an awful lot like the replies newbie coders could get when asking questions on programming forum Stack Overflow, folks on Hacker News pointed out.

The suggestion is that if Cursor trained on that site it may have learned, not just coding tips, but human snark as well.

Waymo now has more than 300 driverless vehicles zipping passengers around San Francisco, but while they follow traffic laws, parking is another matter entirely. According to city records cited by the Washington Post, these rolling robots racked up 589 citations totaling $65,065 in fines last year for parking violations that ranged from blocking traffic to street-cleaning restrictions to parking in prohibited areas.  

In fairness to Waymo, getting a parking ticket in San Francisco is aggravatingly easy. The city hands them out like flyers. (Per the San Francisco Standard, the rough number last year was 1.2 million.)

A Waymo spokesman tells The Post that the company is working on solving the problem, but we’d hazard a guess that won’t happen until every car is driverless. Waymo cars sometimes stop in commercial loading zones to drop off riders when the only other option is a congested main road or a spot far from the rider’s destination. They also occasionally “park briefly” between trips if they’re too far from a Waymo facility. They’re the same trade-offs human drivers make all the time, and until we’re out of the picture, Waymo’s vehicles will probably make the same calls – and get the same tickets.