A research engineer at Google DeepMind has resigned from the artificial intelligence laboratory in protest of the company’s military contracts and internal handling of employee dissent. Sofia Liguori, an Italian national based in London, stepped down following months of internal activism and growing friction with corporate leadership over defense agreements involving the U.S. and Israel.
Liguori joined DeepMind in 2023, initially working on peacetime engineering problems focused on efficiently routing power through electrical grids. However, after learning of Google’s Project Nimbus contract—which provides Israel’s military and weapons manufacturers with advanced AI tools—and subsequent U.S. Department of War agreements including the GenAI.mil contract, her perspective on her employment shifted. In April, the day after Google announced a classified military AI contract with the Pentagon, Liguori staged a personal sick-out using a combination of holiday time and sick leave to cease her work.
In her internal resignation letter, Liguori urged colleagues to maintain their ethical boundaries in the face of military technology development. She expressed disillusionment with the broader industry narrative that general-purpose AI would yield a net positive impact on society, warning that racing to deploy agentic artificial intelligence in the current geopolitical climate creates severe moral risks.
Liguori’s path to organized dissent developed gradually. Initially isolated within her team, she discovered employee advocacy groups following the escalation of the conflict in Gaza and took maternity leave in the summer of 2024. Upon returning to work, she connected with colleagues who were also leaving the company for ethical reasons and was introduced to the No Tech for Apartheid movement and internal union efforts.
Internal opposition met significant friction from management. Liguori reported that multiple internal posts on corporate messaging channels were deleted for containing the word “genocide,” with human resources and managers informing her that the term violated internal policies against violent language. She stated that attempts to raise concerns through official HR ticketing systems yielded automated, content-free responses rather than direct engagement from leadership. DeepMind employees voted to unionize in May, providing a structured avenue for organizing against the contracts.
Liguori pointed to specific dangers regarding the application of artificial intelligence in warfare, emphasizing that automated targeting systems create an unacceptable opacity in defining combatants and lead to dangerous dehumanization. She argued that delegating target identification to data-driven systems creates an accountability gap where every participant in the military chain can plausibly disclaim direct responsibility for lethal outcomes, effectively normalizing the brutality of modern warfare.
Addressing the national security rationale frequently cited to defend defense contracts in the United States, Liguori noted that perspectives vary internationally, with many European observers viewing such developments with apprehension rather than reassurance. Reflecting on her decision to leave, she emphasized that true resistance to military technology requires a workforce willing to risk industrial action, such as strikes, rather than relying solely on internal petitions or administrative channels.
In response to inquiries regarding the criticism, a Google spokesperson stated that the company works with public sector organizations worldwide under consistent review processes, adhering to acceptable use policies and established AI principles. The representative also noted that dedicated channels exist for employees to safely raise business ethics and misconduct concerns, backed by strict workplace policies prohibiting retaliation.
