AI Outperforms Radiologists in Lung Cancer Diagnosis—Insurers’ Black-Box Algorithms Under Fire

The U.S. Department of Energy has chosen the Savannah River Site (SRS) in South Carolina as one of four federal locations for new AI-ready data centers and supporting energy infrastructure.

• Other selected sites include Idaho National Laboratory, Oak Ridge Reservation in Tennessee, and the Paducah Gaseous Diffusion Plant in Kentucky.


• Energy Secretary Chris Wright described the move as part of an effort to "accelerate the next Manhattan Project," highlighting benefits for grid reliability, national security and lower energy costs.

DOE expects to issue solicitations to private partners in the coming months and select developers by the end of 2025.

AI is reshaping both diagnosis and coverage in U.S. lung cancer care. Seattle-based RevealDx has commercialized an AI-enabled software, RevealAI-Lung, that assigns a Malignancy Similarity Index to lung nodules and, in early studies, outperformed radiologists using the 1,800-word LungRads guidelines. A single click now can generate more accurate follow-up recommendations, potentially catching cancers that standard protocols miss.

Meanwhile, health insurers increasingly rely on proprietary AI systems to approve or deny treatment—most notably via prior authorization algorithms that determine "medical necessity." With only 1 in 500 denials appealed and no FDA oversight, critics warn these opaque tools can delay or withhold care, especially from chronically ill and minority patients, prompting calls for uniform regulation and independent testing.

Microsoft has plugged a moderate‐severity misconfiguration in its Copilot Python sandbox after researchers from Eye Security demonstrated a root‐access exploit. In April 2025, the team uploaded a crafted script—landing in /mnt/data/pgrep.py—that manipulated writable paths to execute privileged commands. Although no sensitive data was exposed, Microsoft classified the flaw as moderate, issued a patch in July, and added the researchers to its acknowledgment page. This incident underscores the importance of rigorous isolation controls when deploying AI assistants in enterprise environments.

Federal authorities have tapped an AI-powered Deregulation Decision Tool from the Department of Government Efficiency (DOGE) to pare back roughly 200,000 rules, aiming for a 50% cut by President Trump's inauguration anniversary. The Department of Housing and Urban Development has reviewed 1,083 regulatory sections, while the Consumer Financial Protection Bureau applies the system to all its deregulation actions.

In June, a trial run went awry, as DOGE's AI flagged the VA's internet contract for cancellation due to vague directives and reliance on generic models—errors experts say underscore the challenges of deploying off-the-shelf AI for critical government reforms.

As Georgia students head back for the 2025–26 year, classrooms and campuses look markedly different under new statewide mandates. The Distraction-Free School Act bars phone use during class unless a teacher permits it, prompting Fulton County to require magnetic Yondr pouches and Gwinnett to ban phones for middle schoolers all day.

Safety upgrades include AI-powered Evolv weapons scanners in every high school—and most middle schools—in Fulton, Cobb, Gwinnett and Cherokee counties. Meanwhile, districts from Clayton to Atlanta are unveiling brand-new or fully rebuilt facilities, part of a broader push to modernize aging schools and strengthen security.