One of Meta's top AI researchers, Yann LeCun, is leaving after 12 years with the company to found his own AI startup, he announced. LeCun, who is also a professor at New York University, joined the company in 2013 to lead Meta's Fundamental AI Research (FAIR) lab and later took on the role of Chief AI Scientist. 

LeCun said his new startup would "continue the Advanced Machine Intelligence research program (AMI) I have been pursuing over the last several years with colleagues at FAIR, at NYU, and beyond" and that it would partner with Meta. "The goal of the startup is to bring about the next big revolution in AI: systems that understand the physical world, have persistent memory, can reason, and can plan complex action sequences," he wrote in an update on Threads. "AMI will have far-ranging applications in many sectors of the economy, some of which overlap with Meta’s commercial interests, but many of which do not. Pursuing the goal of AMI in an independent entity is a way to maximize its broad impact."

Speculation about LeCun's future at Meta has been mounting in recent months. Earlier this year, the company invested nearly $15 billion into Scale AI and made the 28-year-old CEO, Alexandr Wang, its Chief AI Officer. Meta also recruited Shengjia Zhao, who helped create GPT-4, making him Chief AI Scientist of its newly created Meta Superintelligence Labs unit. 

LeCun, on the other hand, has been openly skeptical of LLMs. "We are not going to get to human-level AI by just scaling LLMs," he said during an appearance on the Big technology podcast earlier this year. And in a recent talk at a conference, he advised aspiring researchers to "absolutely not work on LLMs," according to remarks reported by The Wall Street Journal.

At the same time, Meta has been reshuffling its AI teams. The company cut "several hundred" jobs from its Superintelligence group, including from FAIR, last month. And LeCun has "had difficulty getting resources for his projects at Meta as the company focused more intently on building models to compete with immediate threats from rivals including OpenAI, Alphabet Inc.’s Google and Anthropic," Bloomberg reported. 

LeCun said he will stay on at Meta until the end of the year. "I am extremely grateful to Mark Zuckerberg, Andrew Bosworth (Boz), Chris Cox, and Mike Schroepfer for their support of FAIR, and for their support of the AMI program over the last few years," he wrote.

YouTube has started a renewed effort to integrate direct messaging into its platform. According to a support page, the service has started testing DMs as a way for users to share and discuss videos. The test is for users aged 18 and up in Ireland and Poland. But while a DM usually comes with some expectation of privacy, Google noted that "messages may be reviewed to ensure they follow our Community Guidelines."

This isn't the video platform's first attempt to provide a messaging angle. YouTube added DMs to its app in 2017, then removed the feature in 2019 in order to emphasize public conversations in comments sections. The new test for sharing within YouTube's ecosystem won't mean any change to other ways you might send people videos. Re-introducing the same system six years after cutting it seems like an odd choice, but Google claims this is "a top feature request," so maybe it'll get a broader adoption this time around.

When Meta announced last year that it was ditching third-party fact checkers in favor of an X-style Community Notes system, the company was careful to note that it would only implement the changes within the United States to start. Now, nearly a year later, the social media company is getting ready to expand the crowd-sourced fact checks to more countries, and is asking the Oversight Board for advice on a potential rollout. 

The company has requested that board weigh in on "factors we should consider when deciding which countries, if any, to omit from the international roll out" of Community Notes. Notably, Meta isn't asking the Oversight Board to advise on the merits of replacing traditional fact-checking organizations. Instead, the company wants guidance on how to approach country-specific challenges and whether there should be any carveouts. 

"We respectfully ask the Board to focus its examination on the country-level factors relevant to omitting countries from the international roll-out, and not on topics such as general product design or the operation of the Community Notes algorithm," the company wrote in its request shared by the board. 

Up to now, Meta has been experimenting with Community Notes on Facebook, Instagram and Threads in the United States only. People who want to be able to author notes still need to be approved, but the company allows anyone to rate notes. However, it seems that the feature so far hasn't gained as much traction for Meta as it has on X. In September, the company said that just 6 percent of the more than 15,000 notes that had been contributed had actually been published. By the end of the month, the number of published notes had risen to over 2,000, according to Meta. More than 90,000 people have signed on as contributors.

In a statement from the Oversight Board, the group said it would consider issues like whether a crowd-sourced fact checking system would make sense in countries with "low levels of freedom of expression" or without a free press, as well as places with "low levels of digital literacy." It also said it was hoping to hear public comments from researchers who have studied different approaches to countering misinformation.

Unlike with a typical case from the Oversight Board, which deals with specific content moderation decisions, Meta has no obligation to implement any of the group's recommendations. But, the company has previously chosen to follow its suggestions in previous policy advisory opinions, including its decision to roll back COVID-19 misinformation rules following a recommendation from the board.

Update, November 19, 2025, 9:15AM PT: Added additional stats about Community Notes adoption.

Roblox is starting to roll out the mandatory age checks that will require all of its users to submit an ID or scan their face in order to access the platform's chat features. The updated policy, which the company announced earlier this year, will be enforced first in Australia, New Zealand and the Netherlands and will expand to all other markets by early next year.

The company also detailed a new "age-based chat" system, which will limit users' ability to interact with people outside of their age group. After verifying or estimating a user's age, Roblox will assign them to an age group ranging from 9 years and younger to 21 years and older (there are six total age groups). Teens and children will then be limited from connecting with people that aren't in or close to their estimated age group in in-game chats.

Unlike most social media apps which have a minimum age of 13, Roblox permits much younger children to use its platform. Since most children and many teens don't have IDs, the company uses "age estimation" tech provided by identity company Persona. The checks, which use video selfies, are conducted within Roblox's app and the company says that images of users' faces are immediately deleted after completing the process.

Roblox didn't provide details on how accurate its age estimation features are, but the company's Chief Safety Officer, Matt Kaufman, said that it was "pretty accurate" at guessing the approximate age of most of its users. "What we find is that the algorithms between that 5 and 25 years old [range] are typically pretty accurate within one or two years of their age," he said during a briefing with reporters.

A spokesperson for Roblox later added that Personas’s models “achieved a Mean Absolute Error (MAE) of 1.4 years for minors under 18 based on testing by the Age Check Certification Scheme (ACCS) in UK.” Parents are also able to adjust their child’s birthday in the app via its parental control settings.

All Roblox users can now voluntarily submit to a face scan or provide an ID to the company to ensure their access to its chat features isn't interrupted. The company says it will be enforcing age checks for all users by January and that people in the Netherlands, Australia and New Zealand will need to comply beginning in early December. Next year, the company also plans to put age restrictions around users' ability to access links to outside social media sites and to participate in Roblox Studio. 

Roblox has repeatedly come under fire for alleged safety lapses even as it's released a flurry of child safety updates in recent years. The company is facing lawsuits from Texas, Louisiana and Kentucky amid accusations that it hasn't done enough to prevent adults from targeting teens and children on its service.

Update, November 18, 2025, 9:45AM PT: This story was updated to add additional information from a Roblox spokesperson.