US State Authorizes AI to Renew Prescriptions Without Doctors - Why This Frontier Exposes the Tension Between Access, Automation and Clinical Accountability

For the first time in the United States, an artificial intelligence system has started renewing medical prescriptions without the direct involvement of doctors. The pilot program in Utah, in partnership with the startup Doctronic, began quietly in December 2025 and was officially announced on January 6, 2026. The initiative allows patients to renew prescriptions for 190 common medications, excluding controlled drugs such as opioids and ADHD medications, paying only US$ 4 per renewal.

At the same time, LinkedIn's 2026 ranking of top professions places the artificial intelligence engineer at the top of the list, with growth of 48% in the number of professionals. In another move, FIFA announced AI-created 3D avatars to assist with offside marking at the 2026 World Cup, while Lenovo and NVIDIA unveiled a billion-dollar partnership for gigawatt-scale AI-optimized data centers.

These 24 hours reveal a fundamental tension: AI is moving from being a promise to taking on critical decisions in health, sport and infrastructure. But at what cost? And with what guarantees?

AI Prescribing Medicines: Gain of Access or Unacceptable Risk?

The Utah program works like this: patients access an online platform, prove residency in the state and answer a clinical questionnaire conducted by the AI. If the answers are within the established parameters, the prescription is sent directly to the pharmacy. Doctronic claims that its system has agreement of 99.2% compared to human doctors in 500 emergency cases.

According to Band report, about 80% of all activity related to medicines in the American healthcare system involves prescription renewals. The Utah government's proposal is clear: to free up medical agendas, reduce queues and give health professionals back time for cases that really require complex clinical judgment.

Margaret Woolley Busse, executive director of the Utah Department of Commerce, said that “the goal is to show that well-regulated AI can improve access, reduce delays in treatment and avoid unnecessary hospitalizations, while keeping doctors at the center of care”. But does this premise hold up in practice?

The Other Side of the Coin: Medical and Legal Concerns

Medical organizations are not convinced. John Whyte, from the American Medical Association, said that “although AI has enormous potential to transform medicine, without the doctor's involvement, it can also pose serious risks for patients”. Concerns include subtle clinical signs that can go unnoticed, complex drug interactions and even misuse of the system by people trying to obtain drugs irregularly.

Another critical point is the absence, so far, of a clear position from the FDA (Food and Drug Administration). Although states regulate the practice of medicine, the agency may take the view that AI systems used to treat or prevent diseases fall under the heading of medical devices, This could lead to additional requirements or even suspend the expansion of the project.

As points out columnist Renato Assis in Estado de Minas, The Utah model is structured with safeguards: express patient consent, full recording of decisions, subsequent human review and exclusion of controlled drugs. He points out that “automating the repetitive is progress; automating discernment is stupidity”.

But there is a fundamental “legal knot”: who is responsible for an adverse outcome of an automated renewal? The supplier, the pharmacy, the state or the doctor? Without a clear answer, there is no legal certainty.

AI Engineer: The Profession That Leads 2026

While Utah tests AI in medicine, the Brazilian job market shows where the money is going. According to LinkedIn survey published by G1, the position of artificial intelligence engineer leads the ranking of top professions for 2026 in Brazil.

The number of professionals in this position has grown 48% in the period analyzed. The average salary is around R$ 8,000, but there are vacancies offering up to R$ 32,000, especially for more experienced professionals. The cities with the most vacancies are São Paulo, Florianópolis and Recife. Of the total number of vacancies, 63.55% are remote, This expands the possibilities for professionals outside the major centers.

What is striking is the concentration of technology-related positions in the ranking: more than half of the 25 positions listed involves the intensive use of technological tools, engineering or data analysis. This reflects a job market that is increasingly driven by information and automation.

But there is a structural problem: despite the expansion, only 10.58% of contracts for the position of AI engineer were women, compared to 89.42% for men. The data points to issues related to access to technical training and the permanence of women in technological careers over time.

Why Does It Matter to Non-Engineers?

Because this concentration of technical talent is directly shaping how companies operate, how governments make decisions and how public services are delivered. In my work with companies and governments, I see that the lack of AI literacy among managers and leaders creates dangerous technological dependency. Without even a minimal understanding of how AI works, organizations become hostage to suppliers and lose the ability to assess risks and opportunities autonomously.

AI is no longer just an IT issue. It is strategic, ethical and political.

World Cup 2026: AI Avatars To Mark Impediments

FIFA has announced that all the players in the 2026 World Cup will be scanned and have an avatar created by artificial intelligence in three dimensions. The main aim is to provide a more accurate response when marking offside and other crucial moves where every centimeter counts.

According to Terra report, The players of all the teams will be scanned by cameras and sensors. An AI computer will create avatars with the exact dimensions of each player. These avatars will be integrated into the automatic offside marking software, significantly increasing its accuracy.

Arthur Hu, director of Lenovo - the Chinese multinational technology provider - said that “no two soccer players are the same. Nobody has the same physique or the same dimensions. So the exact dimensions of each player will be taken into account”.

The images of the avatars in an offside situation (or not) will be broadcast on television to clarify the referee's decision to the public. Mattias Grafström, Fifa's general secretary, commented that “3D avatars with artificial intelligence represent a major advance. Combining accurate player data with advanced visualization strengthens confidence in important decisions and brings fans closer to the sport.”.

Beyond Impediment: AI as Technical Assistant

Fifa and Lenovo also announced a new AI application for data analysis called Football AI Pro, The new system, described as working “like a ChatGPT” for coaches. During each match, cameras in the stadiums and sensors in the uniforms will collect data on the players' performance (mileage traveled, areas occupied on the pitch, decisions after receiving the ball, etc.).

Coaches will be able to consult the AI in real time to support decisions such as substitutions or formation changes. The tool will also be available for post-match use by technical teams and scouts.

Gianni Infantino has predicted that the event will be “the greatest spectacle ever seen on planet Earth”, with an audience of seven million people at the 104 matches in this edition. AI is being used to increase transparency and reduce controversy, These are movements that, if successful, could influence other sports.

Lenovo and NVIDIA: Gigawatt-Scale AI Data Centers

Lenovo and NVIDIA announced at the CES (Consumer Electronics Show) in Las Vegas a joint strategy for deploying data centers optimized for artificial intelligence on a gigawatt scale. The initiative, dubbed Lenovo AI Cloud Gigafactory, was presented as a new infrastructure model designed to accelerate the entry into production of applications based on generative AI, agent AI and high-performance computing.

Second TeleSíntese report, The aim is to meet the growing demand from cloud providers and large corporations for environments capable of processing complex workloads with high energy efficiency, low latency and modular scalability.

During the presentation, Lenovo CEO Yuanqing Yang said that “in the age of AI, value is no longer measured only by computing, but also by how quickly it delivers results”. He pointed out that the gigafactory concept was developed to reduce the time to first token (TTFT), This is a metric used to measure the interval between investment in infrastructure and the entry into operation of AI models in productive environments.

Why have efficiency and speed become critical metrics?

Because the cost of maintaining idle AI infrastructure is prohibitive. The lifespan of advanced chips is around 18 months before new generations make them obsolete. This means that companies need to recover investment quickly or they are left with devalued assets.

The initiative includes support for NVIDIA's next-generation systems, such as the GB300 NVL72, with 72 Blackwell Ultra GPUs and 36 Grace CPUs, and the upcoming Vera Rubin NVL72, with complete integration of processing, networking and storage optimized for AI workloads. The solutions use liquid-cooled architecture and are connected by ultra-low latency networks based on NVIDIA Spectrum-6 Ethernet switches.

Lenovo claims to currently serve eight of the world's top ten public cloud providers. The company also announced that its hybrid AI services will now offer full lifecycle support for these infrastructures, with funding via the as a service (TruScale).

Brazilian AI Elite: 10 Startups Ready to US$ 100 Million

The Brazilian technology ecosystem has reached a new level of maturity. According to a survey by Value Capital Advisors, made at the request of Forbes Brazil, The country ends 2025 with a significant leap forward in the artificial intelligence sector. The number of active startups jumped from 352 in 2016 to 975 companies in 2025. This is an increase of almost 40% in the last half decade alone.

But the market is highly concentrated. The Southeast continues to be the epicenter of innovation, home to 71,18% from operations, The state of São Paulo alone holds 56% of the national market share.

Of the 975 startups mapped, only 23 managed to break through the funding barrier of US$ 10 million. The selectivity of investors is evident: only 16 companies achieved a Growth Score of more than 60 points - a metric that analyzes everything from headcount and market share to a history of M&As (mergers and acquisitions).

From this restricted group, 10 startups have been identified as the “elite of 2026”, with the financial and operational muscle to seek rounds of up to US$ 100 million next year.

Who are the Top 10 Startups?

• Blip: It leads the ranking for the CRM and CX Tech sector. It has raised US$ 60 million and uses Conversational AI and Natural Language Processing (NLP) to scale service for large corporations.
• Nagro: AgFintech that raised US$ 40 million and uses AI to carry out ultra-fast and accurate credit analysis for rural producers, reducing credit release times from 120 days to 48 hours.
• Idwall: Focused on Digital Security, it has raised US$ 38 million. Its differential lies in the use of Machine Learning and Computer Vision for biometrics and fraud prevention in real time.
• Enter: Represents the legal vertical (Lawtech) with US$ 35 million. It invests in specialized AI agents for the automated management of complex legal cases.
• Isa Health: Focuses on the health sector with US$ 30 million raised. Uses Machine Learning models to optimize home hospital care.
• Cobli: Operates in Logistics with US$ 20 million. Uses AI to manage fleets and increase operational safety.
• Traive: Focuses on Agribusiness with US$ 20 million. Its technology is aimed at creating credit risk models for financing the agricultural chain.
• Arvo: Healthtech with US$ 20 million. It uses data intelligence for fraud prevention and auditing in healthcare operators.
• Logcomex: It stands out in Logistics and Comex with R$ 165 million. Focuses on automating the global supply chain.
• Tako: It closes the list as the leading AI HRTech, with R$ 100 million raised. It uses AI agents to revolutionize people management and payroll processing.

For Daniel Lasse, CEO of Value Capital Advisors, the Brazilian AI market is no longer about promises, but about solving critical inefficiencies. “The advance of artificial intelligence in Brazil is beginning to translate into mature business models. In a more selective investment environment, they stand out for combining the practical application of AI and the generation of measurable value,” says the executive.

AI Will Produce “Swarms” of Cyberattacks in 2026

It's not all progress. According to Lumu Technologies, autonomous attacks made possible by AI will establish a new reality, marked by “swarms” of attackers, even more sophisticated malware and the adoption of a strategy based on the presumption of compromise.

According to Terra report, AI agents are going to give a new boost to cyberattacks by using autonomous agents capable of firing 10,000 personalized phishing emails per second, It is also possible to create zero-day attacks instantly and deploy ransomware on thousands of endpoints in less than a minute.

Germán Patiño, vice president of sales for Latin America at Lumu Technologies, said: “The year 2026 will mark a crucial moment: the end of the endpoint-centric security model and a significant shift towards a non-negotiable mentality of assume a security breach. We are no longer debating whether an intrusion will occur, but operating under the harsh reality that it has already happened.”.

Why Does the Presumption of Commitment Change Everything?

Because it changes the logic of security from reactive to proactive. Instead of trying to prevent intrusions, companies need to assume that they have already been invaded and build systems to detect, contain and respond quickly. This requires a profound cultural change and investment in continuous monitoring.

Other trends pointed out by Lumu include:

• More sophisticated malware: By 2026, most intrusions are unlikely to use traditional malware. Instead, malicious software will tend to use AI-generated command strings to orchestrate legitimate system tools (PowerShell, WMI, Python, RMM).
• AI agents take over security operations: AI agents will be responsible for screening alerts, correlating signals and even orchestrating response actions faster and more accurately than human analysts.
• Fierce warfare in the ransomware market: Dominant, well-funded gangs will compete for high-value affiliates and victims, using platform resources, brand reputation and multi-vector extortion.

Recife Uses AI to Prevent Stroke

Not everything is a threat. In Recife, an initiative uses artificial intelligence to detect atrial fibrillation, a type of cardiac arrhythmia that can be an indicator that a person is at risk of having a stroke.

Second VEJA column, The initiative has already analyzed around 4,000 electrocardiograms and detected in 48 people an arrhythmia associated with the risk of stroke. Around 20% of ischemic strokes occur because of atrial fibrillation.

Cardiologist Tieta Albanez, who leads the project in Recife, explains: “The patient doesn't feel anything, but their heart is beating irregularly.” Fibrillation causes the heart to pump blood less efficiently, which favors the accumulation of small amounts of fluid in the left atrium. This leads to the formation of clots (thrombi), which run the risk of breaking loose and circulating through the body.

How does AI fit into this story?

Artificial intelligence is very adept at detecting patterns that may go unnoticed by human eyes. “Diagnosis is by electrocardiogram, but not every doctor feels confident interpreting the tracing,” says Tieta. Estimates used in the project indicate that around 84% doctors report discomfort when analyzing an ECG.

With nine clinical algorithms, the Kardia platform, from a company called Neomed, analyzes each electrocardiogram in about seven seconds and generates a preliminary report. “The technology gives the report. The doctor confirms it,” she summarizes.

Between October 2024 and April 2025, 3,933 electrocardiograms were carried out using the platform in private hospitals in Recife. Of this total, 48 cases of atrial fibrillation were identified. “We have greatly increased the diagnosis of this arrhythmia, which previously went unnoticed,” says Tieta.

Detecting atrial fibrillation early completely changes the treatment. In many cases, protection against stroke comes from the use of anticoagulants, drugs that thin the blood and reduce the formation of clots.

What Do These 24 Hours Reveal About Automation in Critical Decisions?

Let's connect the dots. Utah allows AI to renew prescriptions without doctors. FIFA uses AI avatars to mark offside. Recife uses AI to detect arrhythmias. Lenovo and NVIDIA invest billions in infrastructure. Lumu predicts swarms of automated cyberattacks.

What do these stories have in common? They all represent the frontier where automation meets responsibility. We are no longer talking about AI as an assistant. We're talking about AI as an executor of decisions that affect health, sport, security and critical infrastructure.

In my work with companies and governments, I see three recurring tensions:

1. efficiency versus safety

The promise of speed and cost savings is seductive. But efficiency without safeguards is foolhardy. The Utah model has merit for starting small, with a narrow scope and auditing. But what happens when the exception becomes the rule? And when the pressure for volume and profit outweighs clinical caution?

2. Automation Versus Human Judgment

Algorithms are great for patterns. Bad for exceptions. Medicine - and life - is made up of exceptions. A patient with an atypical history, a rare drug interaction, a subtle symptom that doesn't appear on the questionnaire. AI can miss this. And the cost of missing it could be a life.

3. Access Versus Accountability

Democratizing access to health care is urgent. But who is responsible when something goes wrong? Without legal clarity, innovation becomes institutionalized irresponsibility. States like Utah take the political risk of testing. But the clinical risk lies with the patients. And the legal risk is still open.

Why Does Brazil's AI Elite Matter to This Discussion?

Because it shows that Brazil has the technical and operational capacity to build cutting-edge AI solutions. Startups like Blip, Nagro, Idwall and Enter aren't copying models. They are solving local problems with proprietary technology.

But the ecosystem is still concentrated geographically (71,18% in the Southeast) and with low female participation (10,58% of AI engineers are women). This limits the diversity of perspectives and perpetuates biases.

AI is not neutral. It reflects who builds it. If we continue with homogeneous teams, geographically concentrated and disconnected from local realities, we will build solutions that only serve a slice of the population.

What to do with it all?

First, we need AI literacy at scale. Not so that everyone becomes an engineer, but so that managers, doctors, lawyers, educators and citizens have a minimal understanding of how technology works, what its limits are and how to assess risks.

Secondly, we need regulation that balances innovation and responsibility. The regulatory sandbox model, like the one in Utah, has merit. But it needs transparency, public auditing and clear accountability mechanisms.

Thirdly, we need investment in local infrastructure. We cannot depend forever on external suppliers for chips, data centers and software. Digital sovereignty depends on our own technological capacity.

Fourthly, we need diversity in AI construction. More women, more black people, more indigenous people, more people from different regions and backgrounds. AI that serves everyone can only be built by diverse teams.

In my mentoring and consulting, I help executives and companies navigate these tensions. How to adopt AI strategically, ethically and responsibly? How to balance efficiency and safety? How to build internal capacity without creating technological dependency?

If you're facing these questions, let's talk. AI is no longer about futurology. It's about decisions being made now. And those decisions will define whether we build a fairer future or just a more automated one.

What these 24 hours reveal is that automation in critical decisions is no longer a theoretical discussion. It's a reality. And we need to live up to it.