7 Practical Applications of AI in Government

By combining AI-Media’s LEXI automatic captions with our Alta IP encoder, the broadcaster simplified these complex demands with a scalable end-to-end live captioning solution. While iCap Alta ensured seamless Secure and Compliant AI for Governments caption delivery with next-gen workflows that saved time and reduced costs. Make the most of your live LEXI captions with LEXI Library – our easy-to-use cloud caption archiving and search tool.

In recent years, government agencies have increasingly turned to cloud computing to manage vast amounts of data and streamline operations. While cloud technology has many benefits, it also poses security risks, especially when it comes to protecting sensitive information. To address these challenges, agencies are turning to a secure cloud fabric that can ensure the confidentiality, integrity, and availability of their data in the cloud. In addition to being a solution that can be leveraged today, it is designed to grow with an agency’s cloud consumption and mission focus.

Federal Reserve stands up generative AI incubator

To tackle AI-generated misinformation, model outputs should include watermarks, ensuring citizens can determine when they are presented with AI-generated content. To reduce the chance of bioterrorism attacks, access to systems that could identify novel pathogens may need to be restricted to vetted researchers. To ensure that safety-critical AI systems are built on solid foundations, reducing the chance of accidents, widely used foundation models should be designed with a particular focus on transparency and ensuring they behave predictably. To address cybersecurity risks, systems that can identify novel software vulnerabilities should be used to patch exploits before they are made available to hackers. Second, law enforcement organizations are at a significantly lower level of cybersecurity preparedness compared to the military.

• As a result, it is imperative that policymakers recognize the problem, identify vulnerable systems, and take steps to mitigate risk before people get hurt.
• The past decade has borne poisonous fruit from technological seeds planted before the turn of the century.
• But although it will lead to massive opportunities, this technology is an area that needs clear and significant regulation.

Rather than centrally collecting potentially sensitive data from a set of users and then combining their data into one dataset, federated learning instead trains a set of small models directly on each user’s device, and then combines these small models together to form the final model. Because the users’ data never leaves their devices, their privacy is protected and their fears that companies may misuse their data once collected are allayed. Federated learning is being looked to as a potentially groundbreaking solution to complex public policy problems surrounding user privacy and data, as it allows companies to still analyze and utilize user data without ever needing to collect that data.

Advancing Federal Government Use of AI

The Director shall additionally consult with agencies, as appropriate, to identify further opportunities for agencies to allocate resources for those purposes. The actions by the Director shall use appropriate fellowship programs and awards for these purposes. The report shall include a discussion of issues that may hinder the effective use of AI in research and practices needed to ensure that AI is used responsibly for research. (i)    The Secretary of Defense shall carry out the actions described in subsections 4.3(b)(ii) and (iii) of this section for national security systems, and the Secretary of Homeland Security shall carry out these actions for non-national security systems.

Enable Agencies end-to-end connectivity and visibility across the entire development process from innovation to impact with measurable results. Its features include a Content Library for turn-key compliance obligations and controls, an AI-enhanced Secure and Compliant AI for Governments controls builder, and actionable control task creation and linkages. The system simplifies evidence gathering for control effectiveness, and auto-maps controls to compliance needs, leveraging our AI engine and eliminating manual mapping.

Mitigating the Risks of Using Public AI

Until agencies designate their permanent Chief AI Officers consistent with the guidance described in subsection 10.1(b) of this section, they shall be represented on the interagency council by an appropriate official at the Assistant Secretary level or equivalent, as determined by the head of each agency. Similar discussions must occur in regard to the integration of AI into other applications, but not necessarily with the end goal of reaching binary use/don’t use outcomes. For some applications, the integration of AI may pose such little risk that there is little worry.

The recommendations shall address any copyright and related issues discussed in the United States Copyright Office’s study, including the scope of protection for works produced using AI and the treatment of copyrighted works in AI training. Such actions may include a requirement that United States IaaS Providers require foreign resellers of United States IaaS Products to provide United States IaaS Providers verifications relative to those subsections. According to the costs and implementation time involved in developing an AI solution from scratch, which can cost approximately €160,000 per year and take about a year to complete, it is more advisable for your government department to purchase a pre-built generative AI solution like Typetone.

Microsoft Empowers Government Agencies with Secure Access to Generative AI Capabilities

AI systems are trained on historical data which often contain biased or discriminatory information. CogAbility’s mission is to help public servant organizations empower their staff and the people they serve with AI – safely and responsibly. For example, in a recent study Deloitte estimated that applying AI to government operations should unleash more than $4 billion in labor savings alone. Every day, local agency staff around the country are playing with ChatGPT on their lunch breaks – like 100 million other people do.

Therefore, the push to report the vulnerability is based on the fear that an adversary will either steal or discover the vulnerability as well, and therefore there is a need to patch affected systems before this occurs in order to reduce exposure to the vulnerability. Continuing the EternalBlue example, the NSA is criticized not for using EternalBlue, but rather for failing to report it in order to maintain its usefulness. In the context of an AI system, because the system is already known to be vulnerable but unable to be patched, this tension disappears. This “dual use” nature is not unique to AI attacks, but is shared with many other cyber “attacks.” For example, the identical encryption method can be used by dissidents living under an oppressive regime to protect their communications as easily as it can be by terrorists planning an attack. Different industries will likely play into one of these scenarios, if not a hybrid of both. Autonomous vehicle companies are largely operating under the first “every firm on its own” scenario.

Defining AI, Machine Learning, and Large Language Models

Based on current trends, creating frontier AI models will likely soon cost upward of hundreds of millions of dollars in computational power and also require other scarce resources like relevant talent. The regulatory approach we describe would therefore likely target only the handful of well-resourced companies developing these models, while posing few or no burdens on other developers. Nonetheless, by increasing the burdens to those developing the most advanced systems, the market for such systems may become more concentrated. Governments should therefore subsidize smaller players—for example, via a National AI Research Resource—and wield antitrust powers to address excessive accumulation and abuse of market power. After deployment, companies and regulators should continually evaluate the harm caused by the systems, updating safeguards in light of new evidence and scientific developments. For example, developers should adhere to high cybersecurity standards to thwart attempts by malicious actors to steal their systems.

For digital content like images, these attacks can be executed by sprinkling “digital dust” on top of the target.12 Technically, this dust is in the form of small, unperceivable perturbations made to the entire target. Each small portion of the target is changed so slightly that the human eye cannot perceive the change, but in aggregate, these changes are enough to alter the behavior of the algorithm by breaking the brittle patterns learned by the model. A normal digital image is altered with tiny, imperceivable pixel-level perturbations scattered throughout the image, forming the attack image. While the regular image would be classified correctly by the AI system as a “panda”, the attack object is incorrectly classified as a “monkey”. However, because the attack pattern makes such small changes, to the human eye, the attack image looks identical to the original regular image. Unlike traditional cyberattacks that are caused by “bugs” or human mistakes in code, AI attacks are enabled by inherent limitations in the underlying AI algorithms that currently cannot be fixed.