AI in RA
« AI in RA

#7 — Educational Institutions Adopt AI Policies, EU AI Act Enforcement

August 5, 2026
Recent updates to the EU's AI Act clarify responsibilities across the AI supply chain and set dates for enforcing new transparency rules. In education, institutions are moving to create formal AI strategies as usage among students and faculty becomes widespread, though many schools struggle to support this integration. Meanwhile, new applications are emerging, from AI-powered hypothesis generation for scientific research to tools for managing grants in the nonprofit sector.

AI in Education and Research Institutions

Educational institutions are implementing formal AI policies and strategies, moving beyond individual trials. These initiatives include governance frameworks and professional development across K-12 and higher education. A Digital Education Council survey revealed widespread AI usage in higher education, with 88% of students and 77% of faculty reporting use. However, institutions face challenges integrating AI, leading to unsupported students and faculty underrepresentation in policy creation. This indicates a gap between AI adoption rates and institutional preparedness.

In African higher education, existing AI policies focus on academic integrity and responsible use, though enforcement and impact assessment remain inconsistent. Institutions with structured AI policies generally invest more in training and faculty engagement. A study found variations in AI policy development and readiness across the continent. South Africa, Nigeria, and Rwanda are early adopters, aligning policies with national digital strategies. Most institutions are in aspirational stages due to infrastructure and human capacity limits, and cross-national policy alignment needs stronger regional coordination. Researchers have proposed a phased, ethically grounded AI governance framework for the region.

AI in Research Funding and Grants

European research funders now permit AI use in funding evaluation processes to support administrative tasks. Research leaders, however, caution against AI directly influencing scientific assessment. Governments and nonprofit organizations are exploring AI to address capacity constraints in grants management. Fluxx introduced Finn, an intelligent assistant integrated into its platform, to provide immediate information on grant status, portfolio trends, and operational risks for social good funding.

AI in Scientific Discovery

The U.S. Office of Naval Research plans a "research by AI" initiative, aiming to accelerate basic and applied research by using AI to generate hypotheses. This is part of a broader AI strategy to speed scientific discovery and enhance internal operations. The ONR also intends to deploy specialized AI assistants to support program officers in tracking emerging technologies and synthesizing research findings.

AI Model Developments

July 2026 saw new AI model releases focused on enhancing reasoning and reducing operational costs. Agentic AI continued its presence in real-world applications. Research efforts concentrated on addressing AI hallucinations and increasing the power of smaller models. Large language models enable the creation of sophisticated agents with reasoning, planning, and tool-use capabilities. These LLM-based agents are transitioning from theoretical concepts to practical applications, including autonomous task assistants and complex simulations.

AI Regulation

The EU AI Act Amendment clarifies responsibilities within the AI supply chain. The European Commission's AI Office and national authorities will begin enforcing the AI Act and transparency rules from August 2, 2026. These rules require AI systems to inform users of AI interaction or AI-generated content, including labeling deepfakes and chatbots disclosing their AI nature.

The amendment maintains transparency requirements for most AI systems, effective August 2, 2026, but grants a four-month extension for generative AI. It also softens AI literacy obligations from ensuring to supporting development and expands prohibited AI systems to include those generating non-consensual intimate material. The amendment strengthens the AI Office's powers and addresses how it interacts with other EU laws, such as GDPR.

Additional AI Developments

Sources

  1. AI adoption in education moves to forefront, study says - Installation
    Futuresource Consulting's latest AI in Education report finds 67 per cent of teachers now use AI tools, as institutions introduce governance frameworks and long-term AI strategies across K-12 and higher education. Educational institutions are moving beyond isolated AI trials towards formal policies, professional development and institution-wide strategies. The research found that 67 per cent of teachers now use AI tools, though usage remains concentrated in lesson planning and content creation, pointing to further opportunity for deeper integration into learning activities.
  2. EU AI Act Amendments Defer and Clarify Obligations | Akin
    The EU AI Act Amendment, effective July 27, 2026, extends compliance deadlines for high-risk AI systems by 18 months to two years, while maintaining transparency requirements for most AI systems by August 2, 2026, except for generative AI which has a four-month extension. It expands prohibited AI systems to include those generating non-consensual intimate material and softens AI literacy obligations from ensuring to supporting development. The amendment also clarifies responsibilities within the AI supply chain, strengthens the AI Office's powers, and addresses inter-play with other EU laws…
  3. Commission starts enforcing AI Act rules and new transparency ...
    From 2 August 2026, the European Commission's AI Office, along with national authorities, will begin enforcing the Artificial Intelligence (AI) Act and new transparency rules. These rules require certain AI systems to inform users when they are interacting with AI or when content has been generated or altered by it. This includes labeling deepfakes and requiring chatbots to disclose their AI nature, aiming to reduce deception and provide clear obligations for businesses.
  4. Office of Naval Research Expands AI Strategy to Accelerate Scientific Discovery
    The U.S. Office of Naval Research (ONR) is expanding its AI strategy to accelerate scientific discovery and enhance internal operations, as detailed in its new Science and Technology Strategy. AI and autonomy are primary research focuses, with plans for a "research by AI" initiative to support basic and applied research by using AI to generate hypotheses and accelerate discovery. Additionally, ONR will deploy specialized AI assistants to augment program officers' expertise by tracking emerging technologies and synthesizing research findings.
  5. AI Breakthroughs & News: Complete July 2026 Recap
    July 2026 was a busy month for AI, marked by new model releases focused on improving reasoning and reducing costs, and the continued rise of agentic AI in real-world applications. Investment in AI infrastructure remained strong, while research focused on practical problems like reducing AI hallucinations and making smaller models more powerful. Regulatory discussions continued, emphasizing transparency and data scrutiny, indicating future compliance challenges for businesses.
  6. AI governance in African higher education: Status, challenges, and a future-proof policy framework
    As artificial intelligence (AI) reshapes global education systems, African higher education institutions (HEIs) face pressure to adopt and govern AI ethically and effectively. This study investigates five questions: (1) What is the status of AI governance in African HEIs? (2) How ready are institutions to adopt AI policy? (3) What ethical and operational risks are emerging? (4) How do institutional and national AI strategies align? and (5) What future-proof governance framework can be proposed? Using a desk-based meta-synthesis, the study analyzes over 30 publicly available institutional,…
  7. Fluxx Redefines AI-Powered Grantmaking with Launch of Intelligent Assistant Finn
    Fluxx, a leading provider of AI technology for the social good funding ecosystem, has introduced its new intelligent assistant named Finn. Embedded in the Fluxx platform, Finn offers immediate answers regarding grant status, portfolio trends, and operational risks to program, grant, and leadership teams. This aims to overcome slow and siloed data gathering, marking Finn as the first in a new library of agents for funders.
  8. Building AI Readiness in Grants Management: Where to Start
    Governments and nonprofits face a "capacity crunch" in grants management, leading them to explore AI to support operations. To effectively integrate AI, organizations should start by getting their data, processes, and people ready through standardization and documentation. Early adopters should focus on low-risk pilots like drafting grant narratives to build staff confidence and establish basic AI governance.
  9. AI use to aid grant evaluation 'must not shape scientific decisions ...
    Research leaders have broadly welcomed moves by European research funders to allow use of artificial intelligence in supporting funding evaluation processes, but cautioned about a thin line between that and AI “influencing scientific assessment”. The moves by funders come amid growing use of AI by grant applicants, which is thought to be driving higher application volumes and putting pressure on funders and evaluators.
  10. AI In Higher Education Survey 2026: 88% Student Use
    A new "AI in higher education survey 2026" from the Digital Education Council reveals that 88% of students and 77% of faculty now use AI, indicating its move from novelty to a near-universal classroom tool. Despite this widespread adoption, institutions are struggling to keep up, with many students feeling unsupported and a significant number of faculty lacking meaningful involvement in shaping AI policy. The report highlights a growing gap between AI adoption rates and the preparedness of educational institutions to manage its integration effectively.
  11. Exploring Large Language Model‐Based Intelligent Agents: Definitions, Methods, and Prospects
    ABSTRACT The concept of the intelligent agent represents a long‐standing pursuit in artificial intelligence. Recent breakthroughs in large language models (LLMs) have catalyzed a paradigm shift, enabling the development of sophisticated agents that exhibit advanced reasoning, planning, and tool‐use capabilities across diverse domains. These LLM‐based agents, which leverage natural language as a universal interface for cognition and interaction, are rapidly advancing from theoretical constructs to practical applications, ranging from autonomous task assistants to complex multi‐agent…

Also this week

Full transcript
As AI use becomes widespread in education, schools are moving from scattered adoption to creating formal strategies. That transition from tool to policy is a lead story on AI in RA, where we cover recent developments in artificial intelligence. We begin with this shift in educational institutions. Let's start with regulation, where the European Union's AI Act is now entering an operational phase. Meaning the rules are no longer just on paper. What's the first thing to be enforced? As of August 2nd, the big thing is transparency. The European Commission's AI Office and national authorities have started enforcing rules that require systems to tell you when you're interacting with AI. So, chatbots have to announce they're chatbots, and deepfakes have to be labeled as such. The goal is to cut down on deception. Exactly. But at the same time, a new amendment to the Act is also giving some systems more time to comply. More time for whom? It sounds like they're hitting the gas and the brakes at the same time. High-risk AI systems get an extension of 18 to 24 months, and generative AI gets a four-month extension. But the amendment also expands the list of *prohibited* AI systems. So it's a mixed bag. They're adding things like a ban on generating non-consensual intimate material, but also softening the AI literacy obligation from a mandate to 'ensure' literacy to just 'supporting' it. Now, while regulators are setting up these frameworks, other sectors are dealing with the practicalities. In education, there's a shift from just trying out AI to creating formal institutional strategies. We covered the start of this trend in late July. What are the numbers now? Are teachers actually using these tools? A recent report from Futuresource Consulting found that 67 percent of teachers now use AI tools, mostly for things like lesson planning and creating content. And it's not just teachers. A different survey found 88 percent of students and 77 percent of faculty in higher education are using AI. But as we've been tracking, the institutions themselves are struggling to manage it. There's a gap between these high adoption rates and the schools being ready for it. Many students feel unsupported, and faculty are often left out of policy development. That pattern of high adoption and low institutional readiness seems to be global. A separate analysis of African higher education institutions found a similar landscape of opportunities and challenges. Where policies do exist on the continent, they're focused on academic integrity and responsible use. The study also showed that the institutions that have these structured policies are the same ones investing more in training and engaging with faculty. But it’s very uneven. You have early adopters like South Africa, Nigeria, and Rwanda, where institutional policies are aligning with national strategies. But most other institutions are in what the study calls an 'aspirational phase'. Right, held back by limits in infrastructure and human capacity. In response, researchers have proposed a new governance framework designed specifically for African higher education. We're also seeing AI become more integrated into research and grant funding. The U.S. Office of Naval Research announced an initiative called 'research by AI'. This program will use AI to generate hypotheses and accelerate research. They're also deploying AI assistants to help their own program officers track new technologies. So AI isn't just an administrative tool; it's becoming part of the scientific process itself. We are seeing the administrative use grow, though. European funders have started permitting AI for tasks in their evaluation processes. But with a major note of caution that we mentioned last week. Research leaders are stressing that AI should assist with evaluation, but not influence the final scientific assessment. This is partly a response to a 'capacity crunch' in managing grants. We're even seeing commercial tools for this, like an intelligent assistant named Finn from a company called Fluxx, which provides grant managers with information on portfolio trends and risks. Underpinning all of this are the advances in the core technology. Agentic AI systems are expanding into more real-world applications, a trend that was gaining momentum in July. What's driving that? Is it just more computing power? It's the release of new models that have better reasoning capabilities and lower operational costs. They're enabling more sophisticated agents that can handle complex planning and tool use, moving them from theory to practice. So the engines are getting smarter and more efficient. At the same time, research is tackling persistent problems like reducing AI hallucinations and making smaller models more powerful. A few other things on our radar: US organizations are integrating more AI tools, and more US workers are using them. Though with some apprehension. Many expect AI to negatively affect their jobs, with the benefits concentrated among business owners, executives, and higher-income individuals. Also, the Trump Administration released a plan to renew the U.S. R&D strategy, and a research paper argued for the moral status and welfare protection of conscious AI. And on the technical side, we saw several survey papers published that try to make sense of LLM-based agents—their architecture, how they interact, and how to evaluate their performance. And a final, very specific application: educators developed an AI-assisted dialogue system to help train nurses for multilingual communication with patients. We will have more updates in our next episode. Until then, thanks for listening to AI in RA.

From our community

Builders Group

A peer working group for research administrators who are actively building AI into their offices. Some members work alongside IT teams; others are piecing it together on their own. What unites the group is the work of implementation itself:

RSVP →

From our community

Join AI in Research Administration

Join the community →