**[Meta Ends Collaboration with Dawn Song's Team After 4 Months, Sources Cite "Incompatible Work Styles"]** Beating AI News Flash, according to a report by Semafor, Meta is laying off a group of employees it hired in June from AI safety startup Virtue AI, ending their collaboration after only about four months. Meta spokesperson Andy Stone told Semafor that the layoffs were due to "incompatible work styles," and the partnership did not progress as expected. In June, UC Berkeley computer science professor and Virtue AI founder Dawn Song announced her team’s integration into Meta’s Superintelligence Labs. Joining alongside her were Virtue AI co-founders Bo Li, Sanmi Koyejo, and other team members. The recruitment was reported as Meta bringing in an independent team rather than acquiring the entire Virtue AI company. Virtue AI specializes in AI safety testing, red team attacks, vulnerability discovery, and protection and governance during model operations. Meta’s decision to bring in the Virtue AI team was seen as part of its efforts to advance its AI Agent and superintelligence strategy, as well as to strengthen AI safety and frontier risk governance. According to Semafor, Meta stated that the Superintelligence Labs will continue to focus on AI safety, alignment, and frontier risks. However, Meta did not disclose the specific disagreements between the two parties, nor did it confirm whether Dawn Song and other core members of Virtue AI have all left Meta. As of the time of reporting, Virtue AI has not responded to inquiries. The report suggests that the rapid changes in this collaboration have once again sparked discussions about the issue of "organizational mismatch" when top-tier research teams join large corporations. This refers to the potential conflict between the autonomy, flexibility, and high decision-making efficiency that startup teams are accustomed to and the organizational processes and corporate culture of large tech companies. [Original Link]