Meta's Llama AI Team Suffers Talent Exodus As Top Researchers Join $2B Mistral AI, Backed By Andreessen Horowitz And Salesforce
The tech giant is contending with a major loss of talent within its AI division as the architects behind its flagship Llama models exit the company. According to Business Insider, only three of the 14 researchers credited on the 2023 Llama paper remain employed at Meta.
Five of the Llama paper's co-authors have joined Mistral in the past 18 months, intensifying scrutiny around Meta's ability to retain top-tier AI personnel, Business Insider reports.
Don't Miss:
'Scrolling To UBI' — Deloitte's #1 fastest-growing software company allows users to earn money on their phones.
Hasbro, MGM, and Skechers trust this AI marketing firm —
Backed by $2 billion in funding, Mistral AI is rapidly gaining ground as one of the most aggressive challengers in the open-source AI space. As reported by TechCrunch, the company is supported by premier venture firms including Andreessen Horowitz, Lightspeed Venture Partners, and Salesforce (NYSE:CRM), all known for placing high-stakes bets on transformative technologies. Other notable backers include Bpifrance, Xavier Niel, Eric Schmidt, General Catalyst, and BNP Paribas, reflecting deep institutional and individual confidence in Mistral's long-term vision.
Founded in 2023, Mistral is building advanced open-weight AI models that directly compete with Meta's Llama family. Its approach appeals to developers seeking transparency and customization in contrast to the closed nature of proprietary systems. With former Meta researchers such as Marie-Anne Lachaux, Thibaut Lavril, and Baptiste Rozière now working alongside Mistral's founders, the company may lead the next wave of open-source innovation, Business Insider reports.
Trending:
Meta's previous dominance in this space was largely defined by its decision to release Llama models with open access to their architecture and training data. According to Business Insider, that move helped validate open-weight large language models as viable alternatives to proprietary giants like OpenAI and Google. But with the original architects of Llama now working elsewhere, Meta's early lead is under pressure.
Meta's internal AI leadership is undergoing a shift as well. In April, longtime executive Joelle Pineau stepped away from her role leading the Fundamental AI Research group after eight years. Taking over the position is Robert Fergus, a FAIR co-founder and former DeepMind scientist, marking a return to Meta following a five-year stint at Google's AI lab, Business Insider says.
Separately, The Wall Street Journal reported that Meta's largest AI model to date, dubbed Behemoth, has been delayed due to internal concerns over performance and direction. Meanwhile, Business Insider notes that developers are increasingly turning to faster-evolving alternatives such as Qwen and DeepSeek following the Llama 4 release.Despite investing billions into AI infrastructure, Meta still lacks a model explicitly focused on reasoning tasks, such as multi-step problem-solving or tool use. According to Business Insider, competitors like OpenAI and Anthropic are moving quickly to prioritize those capabilities, and without that strategic leap, Meta's influence over the open-source ecosystem may continue to decline.
Of the 11 researchers who left Meta since the Llama paper's publication, most had been with the company for more than five years, according to LinkedIn profiles reviewed by Business Insider. Some departed as recently as February. Their exits mark a significant shift in Meta's AI capabilities and raise questions about the company's ability to maintain its leadership in the field.
Read Next: Deloitte's fastest-growing software company partners with Amazon, Walmart & Target –
Image: Shutterstock
UNLOCKED: 5 NEW TRADES EVERY WEEK. Click now to get top trade ideas daily, plus unlimited access to cutting-edge tools and strategies to gain an edge in the markets.
Get the latest stock analysis from Benzinga?
SALESFORCE (CRM): Free Stock Analysis Report
This article Meta's Llama AI Team Suffers Talent Exodus As Top Researchers Join $2B Mistral AI, Backed By Andreessen Horowitz And Salesforce originally appeared on Benzinga.com
© 2025 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
20 minutes ago
- Yahoo
Alexander Isak 'rejects second Newcastle contract offer and makes Liverpool demand'
Newcastle United attempted to renegotiate a new contract with Alexander Isak earlier this week, but the player rejected the approach and made it clear that he wants to join Liverpool this summer. That's according to Fabrizio Romano, who claimed during a recent video that Isak has no intention of signing a new deal with Newcastle. Liverpool submitted a formal offer worth $148 million in an attempt to sign Isak earlier this window, but the bid was rejected. Due to uncertainty surrounding his future, Isak has been forced to train away from the rest of the Newcastle squad. Isak is under contract at Newcastle until June 2028, but head coach Eddie Howe confirmed that he will continue to train in isolation unless his attitude changes. READ MORE: Man Utd transfers make a mockery of the attention on Liverpool's spending READ MORE: Liverpool transfer news LIVE: Alexander Isak move, Randal Kolo Muani latest, Marc Guehi decision Liverpool is reportedly ready to submit an improved offer for Isak, but is waiting for Newcastle to allow negotiations. Newcastle has attempted to sign a new striker, and a potential replacement for Isak, but has suffered a disastrous summer transfer window. An attempted contract renegotiation and substantial pay rise for Isak failed, and the player is pushing to join Liverpool. "I'm told that he's saying in a very clear way to Newcastle this week, that he's not willing to sign a new contract at the club," Romano said in a recent update. "Newcastle tried again earlier this week, but Isak has no intention to sign and wants Liverpool." Howe also provided an update on Isak's situation on Friday. "We've had discussions, and it's clear at the moment we can't have him involved with the group," Howe told reporters. "I don't know how long that will be for. "It's clear I can't involve him with the team at the moment. He is training later than the rest of the squad. There have been discussions since the tour. I don't think disciplinary action has happened. "He is here, so of course he has a future at Newcastle. From our perspective, we would love to have him back with the team, but the situation has to be right for that to happen." Newcastle has made several attempts to sign a new striker this summer and was interested in Ipswich forward Liam Delap. Chelsea ended up signing the England striker, and Newcastle began to focus on Eintracht Frankfurt striker Hugo Ekitike. Liverpool then agreed a deal worth $106 million to sign Ekitike, beating Newcastle to the signing. Then, Liverpool began pursuing a record deal for Isak, but Newcastle has yet to enter negotiations. Newcastle then attempted to sign RB Leipzig striker Benjamin Sesko, and finally had a bid accepted. However, Manchester United hijacked the deal after Sesko admitted that he'd prefer to join the Europa League finalists. It's still believed that Newcastle would only allow Isak to leave if a suitable replacement could be signed. However, with Isak unwilling to sign a new deal and his stock high, Newcastle may be forced to sell the striker.
The Hill
22 minutes ago
- The Hill
Meta settles with conservative activist over AI chatbot lawsuit
Meta Platforms settled a defamation lawsuit with Robby Starbuck, who claimed that Meta's artificial intelligence (AI) falsely accused him of participating in the Jan. 6 Capitol riots. There is no publicly available information on the details of the settlement except that Robby Starbuck, a conservative activist working against diversity, equity and inclusion (DEI), will work with Meta to remove 'ideological and political bias' from the company's AI. 'Both parties have resolved this matter to our mutual satisfaction. Since engaging on these important issues with Robby, Meta has made tremendous strides to improve the accuracy of Meta AI and mitigate ideological and political bias,' a joint statement from Meta and Starbuck reads. 'Building on that work, Meta and Robby Starbuck will work collaboratively in the coming months to continue to find ways to address issues of ideological and political bias and minimize the risk that the model returns hallucinations in response to user queries,' he added. Meta did not immediately respond to The Hill's request for comment. Starbuck on Friday told CNBC's 'Squawk Box' that both himself and Meta saw that this problem could impact other users of the company's platforms. 'That was always the point of my lawsuit — is fix this for everybody so this doesn't become a massive, you know, really terrible story in the future where AI affects elections in ways that no one is comfortable with,' he said. Starbuck dodged a question from host Andrew Ross Sorkin about how much money was rewarded, stating that he is still figuring out the details of the collaboration with the tech giant. 'Delivering fairness for consumers is the outcome I've always wanted and I'm pleased to do the work to make that a reality,' Starbuck wrote on social media. 'As we move into a future where AI dominates many parts of our world, now you know that you have an unshakable voice at the table to advocate for ideological fairness.' Starbuck filed the suit against Meta on April 29 and originally demanded more than $5 million from the company. On April 30, Joel Kaplan, Meta's chief global affairs officer, apologized publicly over the matter. 'Robby – I watched your video – this is unacceptable. This is clearly not how our AI should operate. We're sorry for the results it shared about you and that the fix we put in place didn't address the underlying problem,' he posted on the social media platform X.
CBS News
23 minutes ago
- CBS News
Why is Instagram's new Map feature controversial, and should you disable it?
A new feature on Instagram that let's users share their real-time physical location with others on the app has privacy experts concerned over the amount of data exposed and potential safety risks to users. Called Instagram Map, the new feature was introduced on Thursday as part of an app update. On its blog, the company says the location-sharing tool makes it "easier for you and your friends to stay in touch through the content you're enjoying on Instagram." "You can opt into sharing your last active location with friends you pick, and you can turn it off anytime," Instagram said in a blog post announcing the new feature. "You can also open the map to see content your friends and favorite creators are posting from cool spots. No matter how you use the map, you and your friends have a new, lightweight way to connect with each other." In a statement shared with CBS MoneyWatch Friday, Instagram's parent company, Meta, emphasized that Instagram Maps is not automatically active upon updating the app and that users must opt-in to the location-sharing feature in order to make their whereabouts visible to others. "Instagram Map is off by default, and your live location is never shared unless you choose to turn it on. If you do, only people you follow back — or a private, custom list you select — can see your location," a Meta spokesperson said in a statement to CBS MoneyWatch. Users can also choose not to share their locations when they are in particular places, or with particular people. Still, privacy experts say that social media users aren't always aware of how much information they're sharing with an app or its users, even if they have the ability to limit who sees what. "The more these location features are rolled out on social media it carries out the assumption that as long as you give users the ability to toggle them on and off that they'll know exactly how to do that," Douglas Zytko, an app safety expert and associate professor at the College of Innovation & Technology at the University of Michigan-Flint, told CBS MoneyWatch. "But the average user isn't always aware of their privacy settings and if they match their preferences." When the Instagram Map feature is turned on, any content a user posts with a location tagged, including a reel, post or story, will show up on the app's map for 24 hours, according to the Instagram blog. While the feature remains on, the user's location is updated whenever they open the app or return to it. The feature can be turned off at any time. In an Instagram post discussing the feature, Instagram's head, Adam Mosseri, explains how he himself uses the map. "Personally, I use the map to share what I'm up to with a handful of my closest friends, and I curate that list carefully," he said. On Threads, Meta's microblogging site, a number of Instagram account holders claimed that their locations were being pinned on friends' maps by default. Mosseri weighed in, saying the concerns prompted the company to re-examine how the feature works. "We're double-checking everything, but so far it looks mostly like people are confused and assume that, because they can see themselves on the map when they open, other people can see them too," he said. "We're still checking everything though to make sure nobody shares location without explicitly deciding to do so, which, by the way, requires a double consent by design (we ask you to confirm after you say you want to share)." In the top right corner of the app, tap on the messaging function. There you will see a circular world map icon labeled "map. If you click on the icon, you will see you're on location pinned on a map. Friends who are sharing their locations will also appear. Click on the gear icon to choose to share your location with no one, or a custom list of friends, or all of your friends — who are followers that you also follow back on the app. Zytko, however, said it can be complicated for social media users to manage privacy settings that let them share different kinds of content with different groups of people. "This issue is called 'context collapse,'" he said. "Your co-workers see your social media posts, and your friends and family, and there is certain content you only want some groups to see but not others, and it can be hard to manage the visibility of content." Robbie Torney, senior director of AI programs at Common Sense Media, which makes entertainment and technology recommendations for families, said location-sharing features can be particularly risky for younger app users. "These features might feel fun and social, but they create unnecessary risks that teens and many adults don't necessarily understand," he told CBS MoneyWatch. While parents who supervise their teens through controls built into the app can control their kids' location-sharing settings, he still has concerns about the kinds of social pressures such features expose teens, too. Torney said research at Common Sense Media shows that location-sharing also creates "social pressures around where teens go and who they spend time with, and kids feel obligated to share location to show they are someplace cool." Furthermore, when teens share their locations, "they are potentially telling strangers where they are in real time," Torney told CBS MoneyWatch. "If you're not selective about who you're sharing your location with, it creates opportunities for harassment, stalking or worse."
