So give Zuck points for his counterprogramming game, if nothing else. Speaking from an undisclosed location in a vertically-framed video on Tuesday morning, Mark Zuckerberg calmly dropped a tech news bombshell, declaring that Meta would no longer fact-check content posted on its various social media platforms and would no longer restrict posts about certain hot-button topics (more on the news below).

The full implications of this move are still unclear, but it’s safe to say this is a tectonic shift in the tech landscape that will reverberate throughout the business and political worlds (Musk stressed his excitement “to work with President Trump to push back on governments around the world that are going after American companies and pushing to censor more.”)

We’re still two weeks away from Trump taking office, and the pace of change within the tech industry is already dizzying. When it comes to “regularly scheduled programming” for tech in 2025, all bets are off. Here’s the latest news.—Alexei Oreskovic

Want to send thoughts or suggestions to Data Sheet? Drop a line here.

Meta pivots (yet again) on moderation

Meta has performed a massive U-turn on content moderation, days before the second Trump administration begins.

In moves seemingly designed to please Trump and his Republican party, Meta will stop moderating discussions of issues like immigration and gender, scrap its use of automated content moderation for all but “illegal and high-severity” violations, and end its third-party fact-checking program in the US, adopting an X-style “community notes” program instead.

The news was welcomed by Brendan Carr, Trump’s incoming chair of the Federal Communications Commission, who is promising to tackle Big Tech’s “censorship cartel” — Mark Zuckerberg has preemptively capitulated to Carr’s wishes.

Meta is notably not rolling out these changes in Europe just yet, probably because they conflict with the requirements of the Digital Services Act, which requires the likes of Instagram and Facebook to properly tackle disinformation. It remains to be seen how running parallel content-moderation operations will work in practice.

Ironically, just after Zuckerberg and new policy chief Joel Kaplan told the world that Meta was returning to its “roots around free expression,” 404 Media reported that the company has been deleting internal employee criticism of new board member Dana White, a Trump ally. —David Meyer

Anthropic’s big valuation

Claude, how’s your owner’s latest fundraising effort going? Answer: Fabulously.

Anthropic, the company that created popular AI chatbot Claude, is in advanced discussions to raise $2 billion, according to the Wall Street Journal. The investment, if finalized, would value Anthropic at a stratospheric $60 billion.

That would make for the fifth-highest valuation ever for a U.S. tech startup. Only rival OpenAI, Elon Musk’s space cargo business SpaceX, financial company Stripe, and corporate software maker Databricks have higher valuations, according to CB Insights.

Investor appetite for stakes in AI companies has soared over the past two years based on hopes that their technology will be widely adopted by businesses and consumers. So far, however, most of those AI startups—Anthropic included—have been big money losers as they invest heavily in improving their models.

Anthropic’s planned funding, led by Lightspeed Venture Partners, according to the Wall Street Journal, would more than triple its valuation from last year. At that time, Anthropic was valued at a mere $18 billion in a funding round led by Menlo Ventures. —Verne Kopytoff

Nvidia spurs competition for self-driving cars

With GM’s shuttering of Cruise last month, the robotaxi race looked as if it had come down to two major contenders: Waymo and Tesla. 

But Nvidia may have just changed the game with Cosmos, a new platform for “physical” AI systems like robots and self-driving cars, which navigate spaces in the real world.

Training such systems is notoriously difficult, because, as Fortune’s Sharon Goldman writes, “they require massive amounts of hard-to-get data, such as video of every possible humanoid robot movement or hard-to-replicate self-driving scenarios such as snowy roads and car accidents.” That’s one reason companies like Tesla and Waymo, which have vast troves of their own data, have had the advantage.

Cosmos, which Nvidia is making open source, seeks to level the playing field. Nvidia’s technology lets companies train self-driving and robotic systems, using generative AI to simulate different real-world scenarios with synthetic data. Goldman notes that various companies are already using Cosmos, including Uber, Wayve, and Waabi. —AO

More data

—Memphis warns it may not be able to power Elon Musk’s lofty supercomputer plans. Colossus needs juice.

—Getty Images, Shutterstock to merge. Getty CEO Craig Peters will lead the combined business, valued at $3.7 billion.

—Nvidia is now worth more than AMD, Arm, Broadcom, and Intel—combined.

—Dubai billionaire to invest $20 billion in U.S. data centers. The art of the Trump deal continues.

—Illicit content + cartoons = evade detection. People are using AI to make “Minion Gore.” 

—OpenAI ‘losing money’ on ChatGPT Pro subscriptions. $200 per month isn’t enough to offset usage.

—CFTC chair Rostin Behnam to resign. Another crypto regulator steps down ahead of Trump inauguration.

—Spotify sticks to its work-anywhere guns: “You can’t [hire] grownups and then treat them like children.”

Endstop triggered