Google's Gemini Overhaul, Healthcare AI Excels, Compute Surge

Google is signaling a significant strategic shift in its generative AI offerings, with multiple reports on May 17, 2026, highlighting preparations for major Gemini announcements at Google I/O 2026. These forthcoming revelations are expected to include new agentic capabilities and a revolutionary "Gemini Spark" always-on assistant framework, marking a paradigm shift away from traditional operating systems towards an "intelligence system."^[1] This move indicates Google's aggressive push towards persistent AI agents deeply integrated into users' digital lives.^[1] The core advancement lies in Google's commitment to developing AI agents that can proactively execute multi-step tasks across various applications. "Gemini Spark" is envisioned as a foundational element, an always-on framework that integrates AI directly into operating-system-level workflows, inboxes, browsing, and scheduling.^[1] This signifies a move beyond mere chatbot interactions to a more deeply embedded and anticipatory AI experience, where the AI doesn't just respond to commands but anticipates intent and takes initiative. For instance, it could automatically convert a grocery list image into an active delivery order without explicit step-by-step instructions from the user.^[1] Key players in this initiative are Google AI and the broader Android ecosystem, as Gemini-powered Android devices are expected to be at the forefront of this transformation. This strategy directly intensifies competition with other major AI developers, including OpenAI's burgeoning agent ecosystem and Microsoft's Copilot strategy, underscoring a broader industry trend where the battleground is shifting from pure chatbot popularity to workflow dominance and the AI operating layer.^[1] The impact and implications of "Gemini Spark" and these new agentic capabilities are profound. Consumers can expect a more seamless and intelligent digital experience, where their devices and applications work together autonomously to manage tasks and information. For the industry, this represents a significant investment in making AI truly pervasive and proactive, moving beyond isolated applications to a holistic "intelligence system" that can fundamentally reshape user interaction with technology. This focus on persistent, context-aware AI agents suggests a future where digital environments are dynamically managed by AI, aiming to enhance productivity and simplify complex digital tasks.^[1]

Google is poised to unveil a significant expansion of its artificial intelligence capabilities at its annual I/O developer conference, kicking off on Monday, May 19, 2026. The tech giant has confirmed that the keynote will prominently feature the "latest Gemini model updates" and "agentic coding," widely interpreted by industry observers as a strong indication of a Gemini 4.0 reveal. This next iteration of Google's flagship AI model is expected to deliver substantial improvements in multimodal reasoning, seamless integration with Google Workspace, and enhanced agentic reliability.^[1][2]

Beyond core model advancements, Google I/O 2026 is also slated to showcase a deeper foray into AI-integrated hardware. Expected announcements include partnerships for Android XR Glasses with companies like Samsung, Warby Parker, Gentle Monster, and XREAL. These glasses are anticipated to feature a display-free model enabling hands-free interaction with Gemini, marking a potential shift in how users engage with AI in their daily lives. Additionally, Google is on track to launch "Aluminium OS," an Android-based replacement for ChromeOS, further solidifying its ecosystem's integration with AI at the operating system level. This strategic sequencing, following earlier Android platform announcements, aims to focus I/O purely on model releases and hardware innovations.^[1][2]

The anticipated Gemini 4.0 is not merely an incremental update; it represents Google's ambition to significantly advance the narrative in the fiercely competitive AI landscape. Analysts suggest that if Gemini 4.0's benchmarks even match the high performance of competitors like Claude Mythos Preview's 94.6% GPQA score, it would mark a significant narrative victory for Google. The focus on "agentic coding" also signals Google's commitment to empowering developers with tools for building next-generation AI applications that can perform complex, multi-step tasks autonomously.^[1][2] This evolution from phone-based AI to glasses-based interaction and from experimental demos to shipping products highlights a broader industry trend toward ubiquitous and persistent AI assistance, with memory and contextual awareness being critical differentiators.^[2]

Oracle AI has announced a significant breakthrough in machine intelligence, introducing a new inference framework that fundamentally alters how AI systems learn, reason, and generate insights. This development, revealed on May 18, 2026, signals a departure from conventional algorithmic models towards a more dynamic and context-aware form of AI reasoning.^[1] Described as a "historic turning point," this innovation is poised to redefine capabilities across various industries by enhancing training efficiency, inference speed, and contextual understanding.^[1] The core of Oracle's advancement lies in its ability to merge lightweight model efficiency with deep contextual learning. Unlike traditional models that demand extensive computing resources for training, this new framework compresses the training process without compromising the depth of understanding.^[1] It achieves this through the implementation of adaptive feedback loops and optimized data parsing, allowing for more effective processing of nuanced inputs.^[1] This means the system can interpret complex queries, generate creative content, and analyze multi-layered datasets with greater speed and precision, anticipating user intent and adjusting outputs dynamically.^[1] Key players in this breakthrough include the research and development teams within Oracle AI. The company has emphasized that this advancement is still under development, yet early reports confirm its transformative potential.^[1] The new framework is protected by neural architecture improvements and novel training paradigms, addressing long-standing limitations in scalability and contextual awareness.^[1] It enables AI systems to better interpret ambiguous inputs, learn from fewer examples, and adapt across diverse domains, building upon years of machine learning framework refinement.^[1] The impact and implications of Oracle's new inference framework are substantial. For the industry, it suggests a path toward more efficient and powerful AI deployment, potentially lowering the computational barrier for complex AI applications. Users can expect AI systems that behave less like static tools and more like intuitive, responsive learning engines.^[1] This could accelerate the development of more sophisticated AI assistants, advanced data analytics platforms, and highly adaptive generative content tools, particularly benefiting American users navigating a fast-evolving digital landscape where agility is a competitive advantage.^[1]

Anthropic is making notable strides in the development of AI agent architectures with its modular "Agent Skills" standard, which is gaining rapid traction as of May 17, 2026. This new standard enables Anthropic's Claude AI to dynamically load composable "skill packs" for specialized tasks, representing a significant step towards creating more flexible and extensible AI agent systems.^[1] This advancement addresses the need for AI agents to adapt to diverse operational environments and perform a wider array of functions without requiring a complete model overhaul for each new capability.^[1] The "Agent Skills" standard essentially provides a framework for modularity within AI agents. Instead of a monolithic agent, Claude can now integrate specific, pre-defined skill sets as needed, allowing for on-demand expansion of its capabilities.^[1] This could include anything from advanced data analysis and complex problem-solving to interacting with specific software APIs or understanding niche domain-specific languages. This dynamic loading capability enhances the agent's versatility and efficiency, as it only utilizes the computational resources for the skills currently in demand.^[1] Anthropic is the central player in this innovation, building upon its Claude model to foster a more adaptable and robust AI agent ecosystem. This architectural approach signals a move towards more intelligent and resource-efficient AI design, allowing agents to be customized and scaled more effectively. The rapid traction of this standard indicates strong industry and developer interest in modular, composable AI solutions that can evolve and specialize over time.^[1] The impact of Anthropic's "Agent Skills" standard is significant for the future of AI agent development. It promises to empower developers to create more sophisticated and specialized AI agents that are both powerful and adaptable.^[1] For businesses, this means the potential for highly tailored AI solutions that can seamlessly integrate new functionalities as operational needs change, fostering greater agility and innovation. This modularity could also lead to a more collaborative AI development landscape, where different organizations contribute to a library of "skill packs" that can be shared and combined, accelerating the overall advancement and practical deployment of intelligent agents across various industries.^[1]

Meta AI has introduced a significant advancement in large language models (LLMs) with its new SP-KV research, which focuses on a training-time attention mechanism. Highlighted on May 17, 2026, this breakthrough teaches LLMs to proactively predict which key-value (KV) pairs will be required in the future, thereby substantially reducing KV cache memory overhead.^[1] This technical innovation is critical for the effective deployment of long-context and agentic AI systems, addressing a key challenge in scaling generative AI capabilities.^[1] The KV cache is a crucial component in transformer-based models, storing intermediate computations from previous tokens to avoid re-computation during sequential generation. However, as the context window of LLMs expands, the KV cache can consume enormous amounts of memory, becoming a significant bottleneck for efficiency and performance. Meta AI's SP-KV research directly tackles this by enabling the model to intelligently manage this cache, predicting and pruning less relevant information before it becomes a memory burden.^[1] This foresight allows for more efficient resource allocation, paving the way for generative AI models to handle much longer conversations and more complex, multi-step tasks without prohibitive memory requirements.^[1] The key player in this development is Meta AI, demonstrating its continued investment in fundamental AI research. The SP-KV mechanism represents an architectural refinement within the attention mechanism itself, optimizing how models process and retain information over extended sequences. This is not merely an incremental improvement but a core advancement in model architecture designed to enhance the practicality and scalability of advanced AI systems.^[1] The implications of this research are far-reaching, particularly for applications requiring extensive contextual understanding. Developers and enterprises can anticipate being able to build and deploy more capable agentic AI systems that can maintain coherence and relevance over longer interactions, such as advanced customer service agents, sophisticated code generation tools, or highly detailed content creation platforms.^[1] By reducing the computational demands for long-context processing, Meta AI's SP-KV research makes sophisticated generative AI more accessible and sustainable, accelerating the broader industry's push towards truly intelligent and autonomous AI agents.^[1]

The landscape of generative artificial intelligence is undergoing a profound transformation, moving decisively from conversational chatbots to sophisticated, autonomous "agentic AI" systems. This shift is characterized by AI models capable of planning, executing multi-step tasks, recovering from failures, and interacting with various tools and environments without constant human intervention. Multiple analyses from May 17-18, 2026, underscore agentic AI as a defining trend for the year, signaling its transition from experimental demos to reliable production deployments across narrow domains.^{[1][2][3][4][5][6]}

This accelerated embrace of agentic AI comes with a significant increase in computational requirements. NVIDIA CEO Jensen Huang recently revealed that the compute needed for agentic AI has surged by an astounding 1,000% compared to generative AI in just the last two years. This staggering demand is prompting a fundamental buildout of infrastructure, particularly within the energy sector, with major tech companies reportedly accelerating nuclear commercialization and securing dedicated power plants to meet their immense computational needs.^[7] The underlying improvements facilitating this shift include enhanced tool-calling reliability across frontier models, wider adoption of standardized protocols like Anthropic's Model Context Protocol (MCP) for tool reuse, and maturation of observability platforms for debugging complex agentic workflows.^[2]

The impact of agentic AI is already being felt across various industries. Enterprises are moving beyond isolated productivity tools, embedding agents into customer-facing flows and core business operations, such as sales, procurement, and customer research.^[8][9][10] SAP, for instance, has launched "The Autonomous Enterprise," an architecture designed to manage all business resources, with a unified AI platform for building and governing agents to execute core operations.^[9] However, this rapid deployment also highlights a critical human element: the emergence of a new "designer" role. These individuals are not traditional data scientists but rather domain experts who translate nuanced business logic, judgment calls, and unwritten rules into effective agentic systems, emphasizing the need for a hybrid human-AI workforce.^[10]

Multimodal AI, which processes and integrates information from various data types such as text, images, audio, and video, is rapidly becoming the standard for frontier AI models and is demonstrating significant real-world impact, particularly in complex fields like healthcare. A groundbreaking study reported on May 18, 2026, revealed that a novel multimodal AI model, AMIE, outperformed board-certified primary care physicians (PCPs) across 29 out of 32 evaluation axes in simulated telehealth consultations.^[1] This included superior diagnostic accuracy and even consultation-quality metrics like empathy, suggesting a future where multimodal AI could substantially support remote healthcare delivery, pending further real-world validation.^[1]

This breakthrough in medical diagnostics highlights a broader trend where multimodal capabilities are transitioning from a specialized feature to a default expectation in leading AI models. Major players like OpenAI's GPT-5.5 Instant, Google's expected Gemini 4.0, and Anthropic's Claude Opus 4.7 are converging on multimodal input and output, allowing them to understand and generate content across text, image, audio, and video.^{[2][3][4][5][6][7][8]} For example, GPT-5.5 Instant, now OpenAI's default ChatGPT model, boasts improved multimodal reasoning, scoring higher on tests like MMMU-Pro.^[2] Google's Gemini 2.5 Pro already handles text, image, audio, and video, including audio output, with Gemini 4.0 expected to further refine these capabilities.^[2][4][7]

The implications of this multimodal evolution are far-reaching. Beyond enhanced diagnostic tools, it means fewer pipeline hops for developers, as models can natively fuse different modalities for tasks like analyzing news alongside price charts or allowing a field technician to photograph equipment and receive a diagnostic report.^[3][7] The study in healthcare specifically noted that early medical large language models were limited to text-only chatbots, while this new multimodal approach addresses increasing morbidity risks associated with delayed healthcare access, clinician burnout, and an aging global population.^[1] This integration of diverse data streams allows for richer, more accurate analysis and interactive applications, promising to transform various sectors by providing more comprehensive and contextually aware AI assistance.^[9]

As generative AI continues its rapid evolution, the focus on safety, ethics, and regulation is intensifying, with new policies and research initiatives emerging to address the inherent risks. On May 17, 2026, the U.S. Commerce Department's AI Safety Institute (CAISI) finalized pre-deployment evaluation agreements with all five major frontier AI laboratories: OpenAI, Anthropic, Google DeepMind, Microsoft, and xAI.^[1] This critical development signifies that every major AI model will now undergo government evaluation before its public release, establishing a new standard for responsible AI deployment and highlighting a global push for robust AI governance.^[1][2]

Simultaneously, debates surrounding the appropriate level of regulatory oversight for AI tools, particularly in sensitive sectors like healthcare, are also making headlines. On May 17, 2026, it was reported that the Trump administration is proposing to ease certain safeguards governing AI healthcare tools, including requirements for "user-centered design" tests and "AI transparency model cards."^[3] This proposal faces significant opposition from medical professionals and organizations, such as the American Hospital Association (AHA) and the American College of Physicians, who warn that a lack of clarity could undermine clinician trust, increase liability, erode patient-physician relationships, and compromise patient safety by obscuring how AI algorithms are developed and used.^[3]

The broader industry understanding of AI safety is also shifting, moving beyond merely evaluating individual models to a more comprehensive "system safety" approach. This involves layered protections that encompass risk assessment before deployment, rigorous red-teaming and safety testing, implementation of guardrails and moderation layers, human oversight for high-risk decisions, permission controls for AI agents, sandboxing for automated code, and continuous monitoring post-deployment.^[4] Furthermore, research initiatives like the Schmidt Sciences' "Science of Trustworthy AI" Request for Proposals, with a deadline of May 17, 2026, are actively seeking to fund technical research aimed at understanding, predicting, and controlling risks from frontier AI systems to enable their trustworthy deployment.^[5] These efforts collectively underscore a recognition that the future of AI hinges not just on intelligence, but on trust, control, transparency, and accountability within the complex systems it inhabits.^[4]

Dubai Holding, a prominent diversified global investment company, has announced a landmark collaboration with Microsoft to integrate artificial intelligence (AI) at the core of its extensive operations. This strategic move, reported on May 18, 2026, marks the first enterprise-scale AI deployment of its kind across the Middle East and Africa (MEA) region, positioning Dubai Holding at the vanguard of responsible AI adoption.^[1] The initiative signifies a critical shift from isolated AI experimentation to a holistic, organization-wide integration designed to enhance productivity, decision-making, and overall organizational performance across its diverse portfolio.^[1] The collaboration is set to embed AI into Dubai Holding's fundamental processes and decision-making frameworks, promising greater operational efficiency across sectors such as real estate, hospitality, retail, entertainment, and community management.^[1] Employees throughout the Group will gain access to unified AI interfaces, allowing them to directly apply these advanced tools within their daily responsibilities.^[1] A key aspect of this deployment involves the development and implementation of AI agents to automate routine tasks and streamline workflows, fundamentally altering the company's approach to productivity and decision-making.^[1] This aligns with the broader strategic direction of Dubai Holding to leverage advanced technologies as core operational capabilities and contributes to the UAE's ambition to lead in AI adoption.^[1] This enterprise-wide deployment underscores a maturing trend in AI adoption, where companies are moving beyond pilot programs to integrate AI as a core business capability. The partnership with Microsoft builds upon Dubai Holding's existing advanced AI initiatives, including its joint venture Aither with Palantir Technologies.^[1] The immediate impact is expected to bolster Dubai Holding's efficiency across its varied sectors and support wider economic goals for the UAE, including boosting productivity, competitiveness, and long-term growth.^[1] It also reflects an investment in human capital, ensuring employees are equipped to thrive as the technological landscape evolves.^[1] The initiative reinforces Dubai's reputation as a global hub for applied innovation and the ethical implementation of advanced technologies.^[1]

Faraday Future Intelligent Electric Inc. (FF), a California-based global Embodied AI (EAI) ecosystem company, revealed on May 17, 2026, a comprehensive transformation across five key areas: strategy, product, technology and business, finance, capital, and its AI operating system.^[1] Founder and Global CEO YT Jia, who recently returned to the CEO role, outlined the company's commitment to becoming an "AI First" operating and management system, moving towards an "AI-PPTI" model from "PPTIA."^[1] This ambitious shift involves building an AI-native enterprise operating system and a hybrid organization that combines human expertise with AI Agents.^[1]

The strategic transformation redefines FF as a U.S.-based Physical AI ecosystem company, primarily guided by the "AI First" principle.^[1] The company is concentrating on EAI robotics technology, with two main product engines: EAI humanoid and bionic robots, and EAI automotive robots.^[1] On the EAI Device side, FF has increased its 2026 robot shipment target from 1,000 to 1,500 units, focusing on key use cases such as education, security and inspection, reception and guidance, performances, and university research.^[1] The education sector, particularly family education, is anticipated to be the primary use case for the initial phase of the 2C robotics market, with FF aiming to pioneer robotics education products and drive the U.S. EAI robotics education ecosystem.^[1]

This profound organizational overhaul signals a deep commitment to integrating AI into every facet of the company, from governance and decision-making to operations, product development, and execution.^[1] The move is positioned to maximize FF's unique values across three major stages: 2026, the next two years, and the subsequent five years.^[1] The company emphasizes a "zero excuses, zero internal friction, results-oriented" operating mode, reflecting a determination to deliver on its commitments to stockholders and users.^[1] The announcement comes shortly after FF completed a $25 million financing round, providing capital to fuel these transformative initiatives.^[1]

On May 18, 2026, an in-depth analysis highlighted the evolving landscape of cybersecurity in the Gulf region, emphasizing artificial intelligence's transformative, yet dual, impact.^[1] AI is significantly bolstering cyber defense capabilities by improving threat detection, response mechanisms, and intelligence gathering.^[1] However, the same advanced analytical and generative capabilities are simultaneously being leveraged to enhance the sophistication and scale of cyberattacks.^[1] This paradox presents a critical challenge for Gulf states undergoing rapid digital transformation.^[1]

Governments, corporations, and critical infrastructure sectors across the Gulf states - including Qatar, Saudi Arabia, the UAE, Kuwait, Bahrain, and Oman - have actively adopted AI-enabled technologies.^[1] These systems are crucial for detecting cyber threats through anomaly detection, automating responses, and processing vast data volumes beyond human capacity.^[1] This shift moves these nations from reactive cybersecurity models to more adaptive and predictive approaches, vital given the increasing sophistication of malicious actors.^[1] AI's ability to rapidly identify and respond to threats across distributed infrastructures is particularly important for safeguarding critical infrastructure such as government networks, financial institutions, airports, utilities, and energy operations.^[1]

Despite these advancements in defense, the generative capabilities of AI are also empowering cyber attackers to launch more effective and scalable attacks.^[1] AI can generate highly convincing phishing messages, synthetic voice recordings, and large-scale automated deception campaigns, making cybercrime more sophisticated and often more successful than attacks reliant solely on human skill.^[1] A particularly alarming concern is the potential for attackers to manipulate the data used to train AI systems or exploit weaknesses in model behavior, which could lead to compromised cybersecurity systems, distorted outputs, and a reduction in confidence in automated security tools.^[1] This escalating reliance on AI without sufficient human oversight introduces severe risks, necessitating stronger regulation, enhanced skills development, and greater regional cooperation to mitigate systemic vulnerabilities.^[1]

A significant, yet often under-the-radar, development in generative artificial intelligence is the ascendance of Small Language Models (SLMs). In contrast to the massive, resource-intensive Large Language Models (LLMs), SLMs - typically ranging from 1 to 13 billion parameters - are gaining considerable traction for their efficiency, cost-effectiveness, and ability to operate in resource-constrained or on-device environments. This trend is highlighted in various analyses and news items from May 17-18, 2026, positioning SLMs as a critical component for the future of agentic AI.^{[1][2][3][4][5][6][7][8]}

The appeal of SLMs lies in their ability to deliver comparable performance to much larger models on specific tasks, but with dramatically reduced computational demands. For instance, Microsoft's Phi-3 family is noted for matching or outperforming many larger models on language, coding, and math tasks while running efficiently enough for deployment on phones and laptops.^[5] A notable, concrete example of this efficiency was reported on May 18, 2026: the original creator of Redis shipped `ds4`, a tool enabling a 284B parameter model to run entirely on an M3 Max MacBook at 26 tokens/second. This is achieved through 2-bit compression to fit the model into 128GB of RAM and offloading conversation history to the SSD to maintain a 1M token context window, even offering an OpenAI/Anthropic-compatible API for local coding agents.^[9]

The implications of SLM growth are profound for both privacy and economics. Running models on-device or in private environments ensures tighter control over sensitive data, addressing concerns around data residency and cross-border transfer rules, particularly crucial for sectors like healthcare.^[5][6] Economically, SLMs eliminate per-token cloud costs, offering significant savings for businesses and independent developers; some have reported reducing monthly API expenses from thousands to zero by switching to locally run SLMs.^[6] Experts argue that SLMs are not just a niche optimization but are becoming the practical backbone for agentic AI applications, especially in multilingual, regulated, and cost-sensitive scenarios, as they allow for rapid specialization, modular system design, and significantly lower energy consumption and infrastructure costs.

^[5][8]

While enterprise adoption of Artificial Intelligence is reaching near-universal levels, a significant bottleneck is emerging: data readiness. A new AI Momentum Survey from Dun & Bradstreet, though published slightly before the immediate window (May 13, 2026), highlights a critical challenge that remains highly relevant to ongoing developments. The survey indicates that an overwhelming 97% of organizations are actively engaged in AI initiatives, yet a mere 5% report their data is adequately prepared to support reliable, enterprise-scale AI deployments.^[1] This stark disparity underscores the practical difficulties businesses face in moving beyond pilot projects to truly operationalizing AI across mission-critical workflows and systems.

The core issues contributing to this data readiness gap include problems with data access (cited by 50% of respondents), privacy and compliance risks (44%), and fundamental data quality and governance concerns.^[1] Despite these hurdles, businesses are increasingly viewing AI as a critical imperative, with over two-thirds seeing early signs of ROI and more than half planning to increase AI investments in the next year.^[1] This push is exemplified by strategic partnerships, such as Dubai Holding's collaboration with Microsoft, announced on May 18, 2026, to embed AI into the core of its diverse operations across real estate, hospitality, retail, and more. This initiative aims to drive operational efficiency and position Dubai Holding at the forefront of enterprise AI adoption in the Middle East and Africa.^[2]

The drive towards enterprise AI is shifting from isolated tools to intelligent operational systems that support large-scale decision-making and workflow execution.^[3] This includes the deployment of AI agents to automate routine tasks and streamline processes. However, the success of these deployments hinges on robust governance, security, and control frameworks to operate responsibly at scale.^[2] The ongoing challenge is not just about leveraging advanced AI capabilities but ensuring the foundational data infrastructure is clean, interoperable, and governed, preventing a scenario where rapid AI experimentation outpaces the ability to securely and effectively manage the underlying data.^[1][4]

A significant discussion emerged on May 17, 2026, regarding a Trump administration proposal to relax existing safeguards for AI healthcare tools.^[1] The proposal suggests removing requirements governing records and eliminating privacy protections for sensitive medical data, in addition to overhauling standards for data transmission formats.^[1] Proponents of the rule, including the government, argue it could offer clinicians "more health IT choices to meet their needs through increased competition."^[1] However, this move has drawn strong criticism from various healthcare organizations.^[1]

The American Hospital Association (AHA) and the American College of Physicians (ACP) have voiced serious concerns, highlighting the critical role of existing data in fostering trust and ensuring patient safety in AI tools.^[1] The AHA emphasized that current regulations mandate information on how predictive or generative AI applications are designed, developed, tested, evaluated, and intended for use.^[1] They argue that such data are "critical to foster trust in AI tools and ensure patient safety."^[1] The ACP echoed these sentiments, warning that a "lack of clarity could undermine clinician trust, increase liability expense, and erode the patient-physician relationship."^[1] Even developers of AI tools reportedly harbor uncertainty regarding the implications of the proposed changes.^[1]

The immediate implications of easing these safeguards are far-reaching. While potentially increasing choices and competition among health IT providers, the concerns raised by medical professional organizations point to significant risks for patient safety and data privacy.^[1] The proposal's impact on trust in AI-driven healthcare solutions is a central point of contention, with critics fearing that reduced transparency and oversight could lead to misdiagnoses, data breaches, and a deterioration of the patient-physician relationship.^[1] Public comment on the proposal closed in February, and the Department of Health and Human Services (HHS) IT office declined to comment further while the proposal is still undergoing the regulatory process.^[1]