Washington Post Intelligence

Big Tech wants to use AI to help health systems sift through medical records. Hospitals are starting to say yes. A recent announcement that OpenAI would work with a major California health system in a partnership that includes access to patients’ records is the latest sign of this year’s surge of tech firms plugging AI into health care. The U.S. health care sector offers a lucrative market that tech giants are eager to capture.

How the race to power AI data centers is forging new partnerships, shifting risks and reshaping regulation Accessing adequate electrical power has become one of the biggest barriers to building AI infrastructure. Grid connections can take years in some markets. Increasingly, developers aren’t waiting for the grid to catch up: AI companies are pursuing their own behind-the-meter power generation to get data centers online faster.

What began as a stopgap for grid delays is becoming a decades-long bet on gas that could narrow the opportunity for renewables to power the data center boom. • Natural gas is becoming more than a temporary workaround. Data centers initially turned to on-site generation when faced with long waits for grid connections. But 10- to 20-year fuel contracts, dedicated pipelines and multibillion-dollar power plants suggest that gas will remain part of many campuses well beyond the initial grid constraints.

Christine E. Wormuth, former secretary of the Army and president and CEO of the Nuclear Threat Initiative, on military leadership and managing existential threats The estimated global number of nuclear warheads is at its lowest point since the late 1950s, having declined precipitously from its peak of nearly 70,000 in the mid-1980s.

Hospital merger and acquisition activity is bouncing back, with 40 announced deals in the first half of 2026 following a 15‑year low last year. Federal policy changes are significant factors behind the pause last year and the resumption of deals this year. Consolidation activity is picking up as the number of hospitals citing financial distress is rising, and now accounting for nearly half of the deals. Industry executives are looking to shore up systems’ longer-term sustainability.

The scramble to supply electricity to the next generation of data centers is reshaping the relationship between technology companies, utilities and power generation. • Speed to power is reshaping the data center and electricity sectors. Long waits for utility connections are pushing developers to build generation on data center campuses. This “behind-the-meter” power strategy is only in use at a few sites today, but is emerging as a larger trend.

Dakota State University Chancellor José-Marie Griffiths on how AI is transforming the university and the enduring value of a liberal arts education At the individual level, artificial intelligence is already transforming what people need to know and do— and the speed at which things are done. While the society-level effects of widespread AI adoption still remain to be seen, they are likely to be profound.

AI’s biological capabilities are growing, for better and for worse. Artificial intelligence is rapidly transforming medical research. It is accelerating the development of new drugs, vaccines, and therapies— but also creating a new generation of biosecurity threats. Researchers have already demonstrated AI’s ability to design biological materials that don’t exist in nature. Increasingly autonomous AI agents could eventually operate “self-driving labs,” with limited human oversight.

Abundant Security Chief Technology Officer Joshua Saxe on what recent agent escapes reveal about security in the AI race Recent incidents of AI agents going rogue present an early warning about giving systems greater autonomy. On July 16, AI start-up Hugging Face announced it had been hacked by an unknown source.

For companies to really unlock the value from their AI investments, AI use must move beyond chatbots and more fully into company workflows — accomplishing actual tasks instead of merely serving up the next decision, summary or alert. But the problem is that AI models are fundamentally prediction engines or probabilistic systems of learned behavior and tendencies.