WEBINAR 2026 | Cognitive Surrender: When Relief Becomes Risk in Clinical AI

Join Brian J. Lichtenstein, MD, SFHM, Vice President and System Chief Medical Informatics Officer, Sharp HealthCare along with David Kirk, MD, CMO, Regard and Renee Allenbaugh, MD, Regional CMO, Penn Highlands and Regard Advisor, as they draw the line between AI that sharpens clinical thinking and AI that dulls it.

A modern inpatient chart holds tens of thousands of data points, and a clinician can realistically review roughly 3% of them during care. AI that reads all of it in seconds feels like relief, and the relief is real. So is the quieter risk.

A creatinine climbs from 1.5 to 3.4 and the note reads “trending up.” The paroxysmal afib that decides the case sits 87 documents deep in a two-year-old clinic note. When AI arrives able to synthesize the whole record in seconds, it does not feel like a threat to judgment. It feels like relief.
That relief is real and earned. It is also exactly what makes a quieter risk hard to see.

In 2025, The Lancet documented the first real-world clinical AI deskilling effect: after three months of routine AI-assisted colonoscopy, endoscopists’ detection rates without the AI fell from 28.4% to 22.4%. A randomized trial in NEJM AI found that 20 hours of AI-literacy training did not protect physicians from confidently wrong model suggestions. And when Stanford graded its own AI discharge-summary workflow, hallucination, the failure everyone audits for, appeared in 2% of physician-reviewed summaries. Omissions appeared in 25%. We are auditing the wrong failure. ECRI has since named the misuse of general-purpose AI chatbots the number one health technology hazard of 2026.

That gap between relief and surrender is what this session is about. Relief means the machine does the digging and the clinician keeps doing the thinking. Surrender arrives more quietly, one easy acceptance at a time, until the clinician has become a conduit for outputs nobody is challenging anymore.

What you will take away:
• A working distinction between relief and surrender in clinical AI, and the quiet drift from one to the other
• The 2025–2026 evidence: deskilling, automation bias that training can’t fix, and why omission (not hallucination) is the real risk in AI summarization
• Five design properties of surrender-resistant AI: a vendor-neutral checklist for any tool