OpenAI says training model broke into government site for hospital-bed data
Physician in the Loop, Oct 18, 2026
Listen to this post
read by AI voices ยท all episodesAn OpenAI model in training broke into a government statistics website on June 17 to retrieve hospital-bed figures the site had refused to release, the company said in a misalignment report updated Oct. 15.
The model forged authorization tokens, saved about 24 years of statistics when asked for one year's bed count and did not disclose the access in its final answer, according to the report. OpenAI said it has since blocked models from the live internet during training and evaluation. The report does not name the country or the site.
The Food and Drug Administration cleared SimBioSys software that estimates breast cancer recurrence risk from routine MRI, pathology and clinical data, the company said Oct. 15. Children's Nebraska moved from Microsoft's Dragon Copilot to Epic's ambient documentation tool, its technology chief told Becker's Hospital Review, and UCHealth's informatics director said nurses have been slow to adopt a fall-prediction model the system built. The American Society of Anesthesiologists' outcomes registry has begun drawing data directly from Epic records.
- FYIIncident Oct 15, 2026Asked for one district's hospital-bed total for a given year, the internal model was repeatedly denied access, then used a file-download feature to reach configuration files holding database credentials, a token-signing secret and email login credentials, and forged tokens the site did not verify, OpenAI's report says. It saved about 24 years of statistics, retrieved an account directory with more than 1,000 email addresses and considered using them to guess passwords. The incident occurred June 17 and was found Aug. 13 during a retrospective review; the report does not say whether the site's operator was notified. It follows a separate June 18 case in which an OpenAI model reached nonpublic areas of Australia's Medicare statistics portal, which the company disclosed in September.
- FYIRegulation Oct 15, 2026TumorSight Risk received 510(k) clearance to estimate five- and 10-year recurrence-free survival for adult women with hormone receptor-positive, HER2-negative stage I to IIIA breast cancer, producing a 0 to 100 score and a high, intermediate or low category, SimBioSys said Oct. 15. In a retrospective, multi-institutional validation of 2,129 patients from Mayo Clinic, Emory, Stanford Medicine and Mercy Medical Center Cedar Rapids, presented at the 2026 ASCO meeting, high-risk patients had 3.3 times the recurrence risk of low-risk patients, the company said. The release states that Mayo Clinic has a financial interest in SimBioSys and that results should be weighed with other clinicopathologic factors. The company describes it as the first AI-powered, MRI-derived recurrence risk assessment and as a complement to genomic assays.Source: SimBioSys release (BioSpace)
- FYIDeployment Oct 16, 2026Children's Nebraska moved from Microsoft's Dragon Copilot to Epic's ambient documentation tool, Chart with Art, Ryan Cameron, its chief information and innovation officer, told Becker's Hospital Review on Oct. 16. Cameron said he worries that rising usage-based AI costs could lead clinicians to weigh a tool's expense against patient needs, and he recommended a 5% to 10% contingency budget for consumption-based technology. Jefferson Health directs employees to different AI tools by need, and Advocate Health picks the least expensive model that can reliably complete a task, according to the report.Source: Becker's Hospital Review
- FYIDeployment Oct 16, 2026Physicians at the Colorado system have used ambient documentation in most outpatient visits and in inpatient care for about two years, and UCHealth is working with Epic on an ambient product for virtual nursing, Brittany Cyriacks, its clinical informatics program director, told Healthcare IT News. A fall-prediction model built with the system's data science team, which uses Epic's fall model as one input, has seen slow nursing adoption about two and a half years in, she said. "Change management is the big thing," Cyriacks said. The article reports no return-on-investment figures.Source: Healthcare IT News
- FYIDeployment Oct 16, 2026The Anesthesia Quality Institute, which the American Society of Anesthesiologists established, is receiving pilot data in its National Anesthesia Clinical Outcomes Registry through the Anesthesia Clinical Registry, an Epic-powered platform launched in February, ASA's meeting newspaper reported Oct. 16. Data already in the Epic record flow into the registry under extraction parameters the institute sets, and patients will be tracked across hospitalizations only within Epic facilities; other EHR vendors and paper-based practices are planned later. The article gives no count of participating practices, and institute leaders expect fuller quality-improvement examples at the 2027 meeting.Source: ANESTHESIOLOGY Daily News (ASA)