Meta Platforms Inc. Chief Executive Officer Mark Zuckerberg alleged that Facebook was “pressured” by the U.S. government to censor content related to COVID-19 during the global pandemic and that he regrets the company’s decision to accede to the demands.

[time-brightcove not-tgx=”true”]

“In 2021, senior officials from the Biden Administration, including the White House, repeatedly pressured our teams for months to censor certain COVID-19 content, including humor and satire,” Zuckerberg wrote in a letter to the Committee on the Judiciary of the U.S. House of Representatives. And while it was Meta’s decision whether to remove content, he continues, “the government pressure was wrong, and I regret that we were not more outspoken about it.”

During the pandemic, Facebook officials drew ire from critics of lockdowns, vaccines and masking mandates because it removed certain posts, saying they contained misinformation related to the virus or otherwise went against its policies. In all, Facebook took down more than 20 million pieces of content in just over a year. Zuckerberg joins other social media executives, including Jack Dorsey, former CEO of blogging platform Twitter, in lamenting past instances of content moderation that, in their view, went too far.

There’s a growing global debate over how far social media companies should go in policing the comments, images and other content posted by their users. Some platforms believe they should be hands off when it comes to telling users what they can and can’t say online, while some governments say that an overly laissez-faire stance can beget criminal behavior. French officials arrested Telegram co-founder Pavel Durov during the weekend, alleging that the company failed to adequately combat crime on the messaging app, including the spread of child sexual abuse material.

On the eve of the U.S. presidential election, pitting Vice President Kamala Harris against former President Donald Trump, Zuckerberg is also taking pains to appear non-partisan. “My goal is to be neutral and not play a role one way or another—or to even appear to be playing a role,” Zuckerberg wrote in the letter, the contents of which were posted to the Facebook page of the House Judiciary Committee and confirmed by Meta. He was writing in reference to contributions made in the last presidential cycle to support electoral infrastructure.

Instagram parent Meta Platforms Inc. will let users create their own AI-powered chatbots and add them to their profiles, an effort to court creators and further integrate the company’s artificial intelligence software into its most popular consumer products.

The feature, called AI Studio, lets creators with professional accounts make a custom AI chatbot that is “an extension of themselves” and can answer common questions from fans or followers. People can tell their bot what types of questions to answer, or which topics to avoid, Meta wrote in a blog post.

[time-brightcove not-tgx=”true”]

Read More: Mark Zuckerberg Just Intensified the Battle For AI’s Future

The broader universe of Instagram users will be able to make a themed chatbot for their profile, for example a bot focused on mixed martial arts, dining or pets. 

Both features will roll out more broadly in the coming weeks.

Meta first announced AI Studio last September. On Monday, the company started shipping the features for some users, timing that will coincide with Chief Executive Officer Mark Zuckerberg’s appearance alongside Nvidia Corp. CEO Jensen Huang at the ACM SIGGRAPH conference in Denver this week.

Meta has been spending aggressively on AI in recent years, and Zuckerberg has been open about his expectation that the company will develop some of the best chatbots in the world. It already offers its own AI chatbot, called Meta AI, inside Instagram, WhatsApp and Facebook.

Meta released its most recent large language model, the technology used to power chatbots, earlier this month. That new model can solve complex math problems and analyze entire books almost instantly, and also allows users to generate new images of themselves in various scenarios using only a text prompt.