Law firms risk a crisis of 'cognitive surrender' as lawyers treat artificial intelligence as a definitive authority rather than a tool, a new behavioural science study warns. The research finds machine dependence is impairing professional judgment, particularly among junior lawyers using automation for routine work.
The report, The AI Leadership Challenge in Law, was produced by behavioural consultancy Positive Group in collaboration with researchers from Harvard Business School, RSGI and Hubel Lab. It draws on qualitative interviews with 16 senior leaders at commercial law firms, conducted between September and October 2025. Respondents were based across six countries including Australia, and came from firms including Baker McKenzie, A&O Shearman, White & Case, Herbert Smith Freehills Kramer and Gilbert + Tobin.
The study follows high-profile cases of AI-hallucinated outputs reaching courtrooms. It argues firms must move beyond issuing technology guidelines to actively retraining lawyers' cognitive habits.
"Some in the legal sector are potentially sleepwalking into a crisis of judgment by treating AI as a definitive authority rather than a collaborative tool," said Will Marien, CEO at Positive Group. "True risk mitigation isn't found in a compliance policy; it requires a psychological shift."
The pressure to deploy automated tools has accelerated as legal tech valuations climb. The report notes AI legal startup Legora recently reached a US$5.55 billion valuation.
Ben Allgrove, Partner and Chief Innovation Officer at Baker McKenzie, said firms should apply the same scrutiny to AI as to people. "We need to treat AI like a junior lawyer by checking it, questioning it, and training scepticism as a core skill. The risk is machine dependence, so we must balance speed with judgment," he said.
Caryn Sandler, Partner and Chief Knowledge and Innovation Officer at Australian firm Gilbert + Tobin, said training paradigms are shifting. "It's not just about training juniors in legal content, we're also building mindset by interrogating information, questioning sources, and not taking everything at face value," she said. "There's now a focus on continuous improvement, empathy and leadership skills much earlier in their careers."
The study also found polarisation among firm leadership, ranging from extreme hype to outright scepticism, is harming return on investment for AI tools and stalling decision-making.
Recommendations include establishing precise frameworks for the level of human input required on specific tiered tasks, and actively managing the psychological biases that lead to uncritical machine dependence.