
NO FAKES Act clears Senate panel; judge blocks SNAP soda and candy limits; quantum math sharpens cancer AI — Morning Medical Update
Key Takeaways
- Senate-approved NO FAKES provisions would impose up to $750,000 per work for unauthorized AI replicas and extend liability to platforms that knowingly host them.
- AMA leadership argues physicians need enforceable protections against identity and voice misuse, as deepfakes can amplify medical misinformation and market unproven therapies.
The top news stories in medicine today.
NO FAKES Act clears Senate panel
AMA urges Congress to act, citing AI used to impersonate physicians and spread health misinformation.
The Senate Judiciary Committee on June 18 voted unanimously, by voice vote, to advance the NO FAKES Act, sending the measure to the full Senate. The bill (
The American Medical Association (AMA) is urging Congress to pass it.
Federal judge blocks SNAP soda and candy limits
The court found the USDA lacked authority to bar food stamp recipients from buying sugary foods and drinks.
A federal judge on June 22 blocked the Trump administration from enforcing limits on what SNAP recipients in five states can buy with their benefits. U.S. District Judge Amy Berman Jackson
Some said they or family members rely on the affected items to manage conditions such as diabetes and allergies. Jackson wrote that the USDA may approve waivers only for limited purposes, such as improving program efficiency, and that "improving the health and diet of SNAP recipients is not included." The agency has approved food-restriction waivers in 23 states, which Agriculture Secretary Brooke Rollins and Health and Human Services Secretary Robert F. Kennedy Jr. have endorsed as part of the Make America Healthy Again movement, and it said it would keep pursuing the restrictions.
SNAP provides monthly benefits to 42 million low-income Americans.
Quantum mechanics AI improves cancer outcomes
A University of Utah team's method drew two new survival predictors from a small neuroblastoma cohort, outperforming standard biomarkers.
Conventional AI needs far more patient samples than genetic features, which makes it a poor fit for cancer trials that often enroll just 20 to 100 people.
Led by biomedical engineer Orly Alter, Ph.D., the team used the method to pull roughly 6 million tumor and blood DNA and tumor RNA features from a small set of neuroblastoma cases and surface two previously unrecognized predictors of survival and treatment response. The predictors outperformed standard biomarkers, including the MYCN gene now used to guide care, and held up across children treated at different hospitals and times.
Unlike the "black boxes" of neural networks, Alter said, the predictors are interpretable and point to specific genes to target; the team has also validated predictions of adult glioblastoma outcomes and drug targets using CRISPR-Cas9. The work appears in APL Quantum, and Alter has spun the algorithms into a company, Prism AI Therapeutics.





