Archive — US — techcrunch
Maybe AI agents can be lawyers after all
Recent advancements in AI, particularly the release of Anthropic’s Opus 4.6, have seen a considerable improvement in AI agents' performance in professional tasks, including law. Scores increased from under 25% to nearly 30% in one-shot trials, indicating rapid progress. Lawyers may not need to worry about immediate displacement, but confidence in their job security is beginning to wane.
Key points
- Previous assessments of AI capabilities in law showed scores below 25%.
- The new release of Opus 4.6 by Anthropic achieved nearly 30% in trial performance.
- The model includes new features, such as 'agent swarms,' aiding in multistep problem-solving.
- Mercor CEO, Brendan Foody, highlighted the rapid progress, marking a jump from 18.4% to 29.8% in just months.
- Despite advancements, current AI performance remains significantly below human capability (100%).
Why it matters
- Historically, AI has struggled to perform at human-like levels in specialized fields, particularly law.
- Current trends in AI show increasing investment and interest in enhancing model capabilities for complex tasks.
- By 2025, AI trends may further evolve, influencing job markets in legal, corporate, and other professional fields.
While lawyers still aren't at immediate risk of replacement, recent AI advancements signal a shift worth acknowledging, prompting a reevaluation of job security in the profession.