Samwise Healthcare IT Newsletter
Thursday, June 11, 2026
DoD and VA must improve EHR cybersecurity coordination, GAO says
A Government Accountability Office report finds the Federal Electronic Health Record Modernization Office does not fully follow leading practices for cybersecurity and privacy collaboration between the Department of Defense and Veterans Affairs. The shared Oracle Health EHR serves more than 200,000 healthcare providers, a figure expected to exceed 500,000 once VA completes its rollout by 2031. DoD finished its implementation in 2024. A Joint Incident Management Framework, under revision since 2021, remained incomplete as of April. The Change Healthcare cyberattack created a backlog of roughly one million VA prescription claims. GAO recommends both agencies define common cybersecurity goals and performance measures.
Sources: Healthcare IT News
Anthropic wants the government to be able to shut down AI that threatens hospitals
Anthropic has published two policy frameworks urging the federal government to regulate frontier AI more assertively, including granting authority to block deployments posing catastrophic risks to hospitals and critical infrastructure. The Advanced AI Framework proposes mandatory safety testing, independent evaluation, and civil penalties tied to global annual revenue for violations. Anthropic flags cybersecurity as a priority concern for hospitals, noting its Claude Mythos Preview model identified thousands of high-severity vulnerabilities across major operating systems and browsers. A companion Economic Policy Framework addresses workforce displacement and distribution of AI’s financial gains. Anthropic said Congress should not preempt state AI laws unless federal legislation meets or exceeds the proposed standards.
Sources: Becker’s Hospital Review
Health Cyberthreat Sharing Advancing, But Gaps Persist
Healthcare organizations have strengthened cyberthreat intelligence sharing over the past decade, but security programs still lag behind evolving risks, according to GovInfoSecurity citing Jeffrey M. Vinson Sr., a former cybersecurity leader at Harris Health System. Groups like the Health Information Sharing and Analysis Center distribute timely intelligence effectively, yet overall advancements across the sector remain uneven. Many healthcare leaders still treat cybersecurity as a technical function rather than a patient-safety business priority. Smaller and rural providers face the deepest barriers, including limited budgets, staffing shortages, and restricted access to intelligence-sharing networks. The report also examines how artificial intelligence tools may reshape the healthcare threat landscape.
Sources: GovInfoSecurity
The Utah AI physician experiment, explained
Utah’s Office of AI Policy is running a first-of-its-kind pilot allowing Doctronic’s AI to autonomously refill prescriptions for 192 common chronic-condition medications. In Phase 1, physicians manually review every AI decision before it reaches the pharmacy; in later phases, AI would act largely autonomously. In the first months, the AI recommended renewal in 72% of cases, with physicians agreeing 91% of the time. The Utah Medical Licensing Board demanded a halt in April over patient safety concerns; the state declined. Doctronic, which raised $40 million in March, expects to expand to other states. The AMA adopted new policies June 10 to ensure AI never replaces physician judgment in such systems.
Sources: Becker’s Hospital Review
House committee takes step toward blocking Medicare AI prior authorization pilot
The House Appropriations Committee voted unanimously to add an amendment to HHS’ 2027 spending legislation that would block funds from being used to implement CMS’s WISeR pilot or any model that adopts prior authorization in traditional Medicare. WISeR, which launched in six states in January, has drawn bipartisan criticism for delaying care to seniors and adding administrative burdens on providers. Rep. Robert Aderholt (R-AL) called the amendment ‘bipartisan’; Rep. Lois Frankel (D-FL) said ‘prior authorization by artificial intelligence is putting the judgement of doctors and the needs of patients behind.’ The measure must still clear the full House and Senate before becoming law.
Sources: Healthcare Dive
Duke Health launches fund to accelerate nurse-led innovation
Duke Health has established a Nursing Innovation Fund to support nurses in developing, testing, and scaling new ideas to improve patient outcomes and strengthen the nursing profession. The fund was created through a gift from Nancy Schlichting, chair of the Duke University Health System Board of Directors, and her wife Pam Theisen, according to a June 10 news release. Schlichting, a Duke University alumna who graduated in 1976, began her healthcare career at Duke Hospital. The fund is designed to support nurses in imagining and scaling innovations that improve care for patients and families while addressing challenges within the profession.
Sources: Becker’s Hospital Review
OpenAI, Anthropic take steps toward IPOs as healthcare AI footprints expand
OpenAI and Anthropic have both confidentially filed S-1 registration statements with the SEC, signaling potential initial public offerings. Reuters reported OpenAI CEO Sam Altman told staff the company expects to go public within the next year, targeting a valuation of up to $1 trillion. Anthropic, which last raised $65 billion in Series H funding at a $965 billion post-money valuation, has not determined share count or pricing. Both companies have expanded into healthcare: OpenAI launched ChatGPT for Healthcare in January across eight major health systems, while Anthropic launched Claude for Healthcare supporting prior authorization, claims appeals, and patient care coordination. Oura also filed a confidential S-1 in May.
Sources: MobiHealthNews
Why oncology is becoming healthcare AI’s toughest test
Oncology is exposing the deepest limits of healthcare AI, according to Dr. Theepa Dinis, a medical oncologist and clinical informatics analyst at emtelligent. ‘Cancer isn’t one disease, it’s thousands,’ she said. While ambient documentation tools have reduced administrative burdens, they cannot handle oncology’s longitudinal complexity — years of imaging, pathology reports, and treatment histories scattered across unstructured EHR data. AI systems frequently struggle to distinguish events like a 1999 left breast cancer diagnosis from a 2015 right breast cancer recurrence. ‘Most clinical AI fails for a simple reason: the data it’s built on isn’t good enough,’ Dinis said, adding that AI hallucinations in pharmacovigilance present direct patient safety risks.
Sources: Healthcare IT News
Curated by JD · samwise.agency
Leave a Reply
You must be logged in to post a comment.