AI reticence | Partner voice | $81M acquisition, AI glucose monitor cleared, more

News and views you ought to know about: 

Precision medicine company buys AI cancer diagnostic tech for $81.25M. AI-based precision medicine developer Tempus AI is scooping up Paige, another healthcare AI company that works on cancer diagnostics. The buyout has a price tag of $81.25 million, the companies said in a joint announcement. Paige’s technology aids pathologists and researchers in the detection of cancer, where Tempus works on targeted treatment plans—it seems like a good match. Tempus said Paige’s robust datasets were a motivation for the purchase, as they will be used to develop better cancer therapies. The diagnostics company is said to have a unique trove that includes nearly seven million digitized slide images, as well as de-identified, diverse data from people living in 45 countries. 

• What’s being said: “We are confident this partnership is uniquely positioned to maximise and expand the reach of our technology, ensuring it fulfills our mission of delivering powerful, data-driven insights.” - Paige CEO and CTO Razik Yousfi. 
   
• More details: Paige currently has a deal with Microsoft for the use of Azure cloud services. Tempus said it will be responsible for fulfilling that agreement, as the existing agreement will be taken on as part of the purchase.
   
• Did you know? Paige has been around since 2017 and its technology was the first-ever AI tool in pathology to be cleared by the U.S. Food and Drug Administration. 

FDA clears AI-powered glucose monitoring app. The clearance marks a first for an over-the-counter tool in this category. The system, which uses AI to translate fluctuations in blood glucose levels into personalized recommendations for living a healthier life, is developed by Signos. The app works in conjunction with monitors placed on the body, which gather real-time metabolic data. Metrics such as diet, exercise, and sleep habits are taken into account by the AI, all in an effort to support weight loss and overall health for people with diabetes—and those looking to stave it off. It’s available now at Signos’s website. 

• Signos was formed in 2018. Its claim to fame is its proprietary continuous glucose monitoring AI, used in this FDA-cleared application. 
   
• The app can be used on a smart device, and it costs between $129 and $139 a month. All necessary glucose monitors are provided by the company, with the cost included in the monthly subscription.

In Illinois, Meta AI is not allowed to be your therapist. The state has banned the use of AI in mental and behavioral healthcare, outside of deploying it in an administrative capacity. The Wellness and Oversight for Psychological Resources Act (HB1806)—signed into law by Gov. JB Pritzker—forbids anyone to use AI to provide mental health treatment, such as through a chatbot, and additionally bans its use in “therapeutic decision-making,” an announcement from the governor’s office stated. Licensed clinicians are still free to use the tech to support patient care, as long as any actual treatment and counseling is conducted by a real person. The state passed the law, arguing it was necessary to protect consumers from low-quality AI apps that offer algorithm-driven mental health counseling and do little more than mine for data. Violators will be hit with steep $10,000 fines.

• Lawmakers argue that the Wellness and Oversight for Psychological Resources Act will also protect the jobs of licensed, trained behavioral health therapists whose positions are threatened by easy-to-access smartphone apps run by cynical profiteers. 
   
• The Illinois Department of Financial and Professional Regulation, lawmakers in the Illinois General Assembly, and the National Association of Social Workers, Illinois Chapter, all had a hand in crafting the initial bill. 
   
• The law went into effect immediately after being signed on Aug. 4.

GPT-5 is less of a pushover. Changes have come to ChatGPT with the launch of 5.0. One of the top concerns since 4.0 has been the system's tendency to be agreeable and sycophantic, particularly when asked to judge something a user submitted. While 5.0 was released with the intention of being less flowery and personal with its language, that didn’t entirely solve the problem. Instead, it just made users on the unofficial Reddit forum feel as if they’d lost a friend. Developer OpenAI may have found a middle ground. Earlier this month, ChatGPT was loaded with adjustable “personalities” that users can cater to themselves—meaning the AI will be more or less overtly flattering, based on individual preference. 

• OpenAI said GPT-5 has been given four distinct personas: Cynic, Robot, Listener and Nerd. The company said all four are equipped with an adjuster bar to reduce sycophancy. These were officially released on Aug. 10, but more updates and tweaks are likely on the way. GPT-5 has seen a steady stream of modifications since launch, most of which are a direct response to feedback from users. 
   
• The company maintains that 5.0 worships users less often than 4.0: "Overall, GPT‑5 is less effusively agreeable, uses fewer unnecessary emojis, and is more subtle and thoughtful in follow‑ups compared to GPT‑4o," OpenAI wrote. "It should feel less like 'talking to AI' and more like chatting with a helpful friend with PhD‑level intelligence."
   

From AIin.Healthcare’s sibling outlets:
 

• Radiology Business: Imaging-guided surgery startup Method AI raises $20M in Series A financing
   
• Health Imaging: AI-enabled devices lacking validation data prior to FDA clearance more likely to be recalled

The ambient AI playbook: Lessons from two leading health systems

At the recent CompassionIT Summit, leaders from Akron Children’s Hospital and Denver Health shared powerful lessons from rolling out ambient documentation to over 1,500 clinicians. Their biggest takeaway? Stories, not stats, drive adoption. Whether it was a heartfelt testimonial that swayed an entire department or a 60-second Nabla demo that eliminated training anxiety, the common thread was simplicity, authenticity, and clinician-centered design. Read more about the way these health systems are navigating ambient AI implementation: https://dhinsights.org/news/the-ambient-ai-playbook-lessons-from-two-leading-health-systems

Tech vendor Reveleer and policy analytics firm Mathematica released their 2025 State of Technology in Value-Based Care report, the findings of which show that while AI adoption in healthcare is widespread, only about 40 percent of organizations said they are “fully committed” to integration that places the technology at the center of all business operations.

In surveying both providers and payers, Reveleer and Mathematica sought to better understand the disconnect between stated support for value-based care models and the logistics of AI adoption that would make the transition away from fee-for-service a reality.

Here are three key takeaways from the survey, based on responses from over 200 directors, executives and C-suite leaders.

1. AI adoption is widespread, but fears of the future run deep.

Looking broadly at the results, both payer and provider groups overwhelmingly believe in the power of AI tech and data; however, they report that data fragmentation and weak collaboration have stalled value-based transitions. 

AI adoption was a particular focus. Despite 93% of providers and 97% of payers increasing their use of AI last year, deep concerns remain around algorithmic transparency, the logistical challenges of scaling infrastructure and staff training. These hangups have left adoption siloed—limiting AI use to one or two tasks, such as improving operational efficiency or supporting physician decision-making.

"Industry leaders widely acknowledge that data provide a competitive advantage, but siloed systems and quality challenges continue to block its full utility. Payers and providers broadly adopt AI, but trust gaps, limited training and governance shortfalls restrain its transformative power."

2. Fragmented data and a lack of interoperability erode confidence in value-based initiatives.

There was near-universal support for the strategic value of data in improving healthcare delivery—from patient outcomes to controlling costs. However, only about a third of organizations seem ready to fully utilize it. Roughly a third of respondents rated their data integration capabilities as “excellent,” and fewer than 50% said they were confident they had enough complete data sets at their disposal to scale value-based care models.

The report’s authors—Reveleer, Mathematica, and Harris Poll—wonder whether the situation would improve if providers and payers had access to each other’s data. While nearly all payers (97%) and providers (100%) agree on their value-based care goals, they are not yet collaborating effectively.

"Part of the challenge is that providers and payers see different parts of the data picture. Payers often have access to a broad set of population data, while providers track detailed, patient-level data. But when supported by modern data infrastructure and the right technology, these two perspectives combined can shed light on what’s working today and uncover new innovations, ready to scale."

3. Technology isn’t enough—providers, clinical staff, patients, and payers are just not ready.

While AI is being widely adopted and data is at the center of healthcare operations, the report underscores that technology alone isn’t enough when investment in people is lacking. Providers and payers alike said gaps in staff training, combined with a lack of AI governance, limit understanding of how technology can be used to make value-based care a reality.

The lack of training also adds to concerns around data safety and cybersecurity, where human error still tends to be a primary source of data breaches.

“[Organizations must empower] and train staff: Move beyond basic introductions to technology by offering extensive, hands-on training tailored to both clinical and administrative roles. Encourage staff to harness technology to make daily tasks more efficient, while cultivating change champions who drive adoption and ensure compliance with data and security standards.”

The full survey is available here.