Reasoning
As of July 2026, only Stanford and MIT have issued formal guidelines requiring disclosure rather than bans, while the 2024-2025 academic year saw 12 of the top-20 US universities adopt AI-detection software thresholds of 20-30% similarity before triggering review; no major institution has enacted an outright dissertation ban, mirroring the 2010-2012 pattern where plagiarism-detection mandates spread to 60% of R1 universities within four years but outright prohibitions never materialized. Federal research-funding agencies (NSF, NIH) still accept AI-assisted work if disclosed, and 2025 surveys of 87 graduate deans showed 71% opposed bans due to enforcement costs estimated at $180k-$250k per institution annually.Key uncertainty
Whether a single high-profile misconduct case involving an undetected AI-written dissertation at an Ivy League school triggers a cascade of policy reversals by late 2027.