Amrit Singh Bedi, an assistant professor in UCF’s Department of Computer Science, has received a one-year, $340,000 award from the Defense Advanced Research Projects Agency, or DARPA, to investigate how artificial intelligence systems can become more resilient. Bedi and his research team will study methods for training AI to withstand failures, adversarial attacks and unexpected conditions. Their central objective is to develop systems capable of identifying weaknesses in their own reasoning and adjusting their behavior over time, rather than depending entirely on fixed safety rules or human intervention.

The project addresses a growing technical problem in modern generative AI. Large language models and related systems are trained on enormous collections of data and optimized to produce useful responses, but their apparent fluency does not guarantee reliability. These models generate outputs by identifying statistical patterns and selecting likely sequences of information. As a result, they may confidently produce false statements, follow misleading instructions or make decisions based on subtle cues that have nothing to do with the intended task. Their performance can deteriorate even further when they encounter inputs designed specifically to exploit their limitations.

Such attacks are often known as adversarial prompting or prompt injection. An attacker may introduce instructions that conflict with a system’s original goals, conceal malicious commands within ordinary text or manipulate the order and wording of information presented to the model. In an AI agent connected to software tools, databases or physical equipment, a successful attack could have consequences beyond a wrong answer. It might cause the system to reveal sensitive information, execute an inappropriate action or abandon the objectives established by its developers. Bedi’s research is focused on helping AI detect these pressures and respond without losing control of its mission.

The SafeRR AI Lab, led by Bedi and his collaborators, will explore what the researchers describe as “self-hardening.” The concept is comparable to strengthening a computer system after identifying a vulnerability, but it is intended to operate within the AI’s reasoning process. A self-hardening system would monitor its own responses for signs of manipulation, inconsistency or unusual uncertainty. When it detects a suspicious instruction or a conflict between competing goals, it could slow down, reassess the evidence, reject the unsafe direction or choose a more conservative response.

Developing such capabilities requires more than simply adding another list of prohibited words or behaviors. Static safeguards can be effective against known problems, but they may fail when an attack is expressed in an unfamiliar form. The UCF team is instead examining adaptive approaches in which an AI system learns from attempts to deceive or derail it. This could involve testing a model against hostile inputs, analyzing where its reasoning breaks down and updating its decision-making strategies to handle similar situations in the future. The challenge is to improve resilience without making the system so cautious that it becomes unable to perform legitimate tasks.

At the center of the project is the distinction between apparent capability and true robustness. An AI may achieve impressive results under normal testing conditions while remaining fragile when instructions are ambiguous, data are incomplete or an adversary is actively trying to manipulate its behavior. For systems used in mission-critical environments, that gap can be dangerous. A small deviation in an autonomous platform, a cybersecurity tool or a defense-related decision-support system could produce consequences far greater than the original error. Bedi’s team aims to develop evaluation methods that reveal how systems behave under pressure, not only how well they perform in ideal conditions.

The research could have implications well beyond military applications. AI tools are increasingly being considered for healthcare, emergency response, robotics, transportation and cybersecurity, where they must operate in environments that are unpredictable and sometimes hostile. A medical system may encounter incomplete patient information, an emergency-response platform may face rapidly changing conditions and a robotic system may receive conflicting signals from its surroundings. In each case, the ability to recognize uncertainty, preserve core objectives and recover from a faulty decision could be as important as the ability to generate a correct answer in the first place.

Bedi describes the broader goal as building AI that can actively recognize, withstand and recover from adversarial pressure while remaining faithful to its intended purpose. If successful, the project could contribute to a new generation of systems that do not merely follow safety instructions but participate in maintaining their own reliability. The researchers are not suggesting that AI should operate without oversight. Instead, self-hardening is intended to provide an additional layer of protection, allowing systems to identify risks earlier and respond more responsibly before human users must intervene. As artificial intelligence moves into increasingly consequential roles, that capacity for self-monitoring and adaptation may become essential to earning public and professional trust.

Subject of Research: Artificial intelligence resilience, adversarial robustness, self-hardening AI, machine learning safety

Article Title: UCF Researchers Explore Self-Hardening Artificial Intelligence to Resist Attacks and Failures

References: Defense Advanced Research Projects Agency grant; University of Central Florida SafeRR AI Lab

Keywords

Artificial intelligence, generative AI, machine learning, adversarial attacks, prompt injection, AI safety, resilient AI, adaptive systems, robotics, cybersecurity, defense technology

The complexity of AI regulation is underscored by the challenges inherent in balancing innovation with risk mitigation. China’s multifaceted approach recognizes that fostering technological advancement cannot come at the expense of safety or ethical responsibility. The draft frameworks currently under expert review emphasize a nuanced blend of exemptive laws, expedient judicial processes, and adaptive requirements that prioritize experimentalism, enabling developers to test and iterate rapidly within safe and regulated boundaries. This signals a shift from rigid regulation to a more flexible and pragmatic oversight regime capable of evolving alongside the technology itself.

Central to China’s regulatory vision is the ambition to mitigate extreme risks associated with AI deployment, including potential threats arising from unregulated autonomous systems and misuse of AI capabilities. The emerging draft laws articulate safeguards designed to preempt and respond to these hazards, embedding mechanisms for risk assessment and ensuring accountability. This proactive stance reflects a growing recognition that AI, while a powerful tool for societal advancement, necessitates robust governance frameworks to prevent harm and maintain public trust in technological progress.

The ongoing legislative process is deeply informed by two key proposals: the Model AI Law and the AI Law, also called the Scholar’s Proposal. These documents encapsulate comprehensive policy ideas that address technical, ethical, and regulatory aspects of AI governance. They are instrumental in shaping parliamentary dialogue, serving not only as a basis for legal drafting within China but also as influential models that may inspire regulatory approaches globally. Their detailed provisions cover core issues such as transparency, data governance, algorithmic fairness, and user rights, laying the groundwork for responsible AI stewardship.

An essential feature of China’s regulatory approach is its stress on maintaining openness and collaboration. Recognizing the inherently global nature of AI development, Chinese policymakers emphasize creating an environment where domestic developers can innovate freely while engaging constructively with the international research community. This approach seeks to harmonize national innovation agendas with global standards, enabling knowledge exchange and joint efforts to address the transnational challenges posed by AI, particularly in domains such as AI safety and ethical compliance.

The regulatory pillars advanced in China reflect a strategic intent to streamline governance rather than fragment it. Previous attempts globally have often stumbled due to inconsistent or overlapping regulations that stifle innovation and obscure accountability. By consolidating AI oversight under a coherent framework that unites legal clarity, responsive judiciary mechanisms, and adaptability to technological evolution, China aims to craft a more efficient and transparent operational environment. This holds promise not only for domestic industry but also for multinational enterprises engaging with Chinese AI markets and research institutions.

Moreover, China’s regulatory model is notable for its embracement of experimentalist governance practices. Rather than imposing rigid top-down mandates, the approach encourages iterative policy adaptation informed by real-world application and multi-stakeholder feedback. This dynamic regulatory mode is particularly suited to the fast-paced AI domain, where technologies evolve rapidly and unforeseen risks may arise unexpectedly. By embedding flexibility, the framework endeavors to be both resilient and innovative, fostering a regulatory culture that learns and adapts continuously.

Global AI governance faces numerous challenges, including geopolitical tensions, variations in legal traditions, and differing societal values. China’s emerging regulatory position thus enters a complex international context where coordination and harmonization are difficult yet essential goals. Experts highlight that China’s framework, by championing safety and openness, offers a constructive avenue for dialogue and partnership among nations. As policymakers worldwide grapple with AI’s dual-edged potential, China’s model may act as a catalyst for cooperative efforts toward shared standards and mutual oversight.

Technically, the proposed regulations address critical components such as algorithmic transparency, mandatory risk assessments, and enforcement provisions. The Model AI Law and Scholar’s Proposal prescribe detailed frameworks for monitoring AI systems’ behavior, ensuring they meet predefined safety criteria and comply with ethical standards. They also underscore the importance of data governance—mandating stringent controls over data quality, privacy, and security—to prevent biases and misuse. These technical pillars reflect cutting-edge understanding of AI’s operational risks and governance needs, aligning with international best practices.

In terms of adjudication, China is considering mechanisms for expeditious legal resolution related to AI disputes, recognizing that the technology’s complexity requires specialized judicial expertise and streamlined procedures. Efficient adjudication serves a dual purpose: it enhances regulatory certainty for innovators and ensures timely redressal of harms, thus safeguarding public interest. This intention to integrate judicial efficiency within the AI regulatory framework illustrates China’s commitment to create a balanced ecosystem where innovation can flourish within legally defined boundaries.

China’s growing influence in open-source AI development further magnifies the significance of its regulatory trajectory. As Chinese AI models and tools become embedded in global AI research infrastructure, the norms and standards China cultivates will inevitably ripple across borders. Hence, Chinese lawmakers’ decisions today not only shape their domestic landscape but also contribute substantively to shaping global AI governance paradigms. The international community watches closely as China crafts legislation that could either facilitate or hinder global efforts for AI cooperative safety management.

While the evolving Chinese AI regulation landscape holds promise, it also calls attention to the need for greater clarity and simplification. Current fragmented policy proposals must coalesce into a unified legal framework capable of clear communication to all stakeholders, from developers to end-users. Achieving such clarity, alongside harmonization with international regulatory trends, will be pivotal in ensuring the framework’s effectiveness and adoption. The coming years will be crucial as China consolidates its legal statutes and enacts binding legislation that operationalizes these ambitions.

In conclusion, China’s strategic approach to AI governance marks a significant chapter in the global narrative of managing transformative technologies. By integrating exemptionary laws, experimental governance, risk mitigation, and global cooperation, the country presents a comprehensive blueprint aiming to balance innovation and responsibility. The decisions emerging from the Chinese legislature will influence not only the domestic innovation landscape but also the broader quest for global AI safety and ethical standards—an endeavor whose stakes have never been higher in the rapidly unfolding digital age.

Subject of Research: Emerging Artificial Intelligence Regulation in China and Its Implications for Global AI Governance
Article Title: China’s emerging AI regulation towards an open future for AI
News Publication Date: 9-Oct-2025
Web References: 10.1126/science.ady7922
Keywords: Artificial Intelligence regulation, AI governance, China AI law, Model AI Law, Scholar’s Proposal, AI safety, innovation policy, global AI collaboration, risk mitigation, algorithmic transparency, experimentalist governance, AI adjudication