The White House is creating a new Artificial Intelligence cybersecurity clearinghouse to coordinate measures for protecting critical infrastructure. Industrial sectors and owners of critical infrastructure are rushing to adapt to new AI models, which possess increasingly complex capabilities for detecting and exploiting vulnerabilities in cybersecurity, as well as for defending against them.
Some companies developing AI have even decided to delay the wide release of their most advanced models to give key partners time to fix identified vulnerabilities before mass distribution. The White House center, named Gold Eagle, is a joint project of the Department of the Treasury, the Department of Homeland Security, and the Pentagon. The platform will be used by AI and cybersecurity companies, as well as critical infrastructure providers such as banks and utilities, for information exchange and coordination of efforts.
According to a high-ranking White House official, the new capabilities allow for vulnerability detection on previously unseen scales. The goal of the new center is to prevent conflicts and ensure that resources are not wasted on fixing or scanning the same vulnerabilities, as well as validating them. Afterward, a team of industry and government engineers handles the prioritization, sorting, and remediation of these vulnerabilities.
The White House declined to disclose the names of the companies participating in the project, referring to them as 'open-source software partners and American companies owning critical infrastructure.' Open-source software has publicly available source code, allowing anyone to freely view, use, modify, and distribute it. In contrast, the most well-known AI models from companies like OpenAI, Google, and Anthropic are closed-source.
Although open-source software is widely distributed, it is often managed by volunteers who may lack the funds and time to ensure the code's security. In 2021, a critical flaw in such software made hundreds of millions of devices worldwide vulnerable to hacking, prompting an active response from officials in the Biden administration.
The center, announced on Tuesday, was established under an executive order signed by President Donald Trump in June. This order also requires AI development companies to provide advanced models to the federal government for review 30 days before their release to trusted partners. The framework for this regulation is expected to be established by early August, but it has not yet been made public.
Nevertheless, the White House has already restricted the release of new AI models in other ways, such as imposing export controls on Anthropic, which were later lifted. Furthermore, the White House asked OpenAI to limit the release of its latest model. This seemingly inconsistent policy regarding the regulation of new model releases has led to widespread calls from the AI industry for more consistent regulation.
A new attack methodology has been identified, exploiting an inherent limitation in artificial intelligence (AI) models and having the potential to compromise thousands of computers. According to Ars Technica, this discovery affects nine popular tools used by programmers, demonstrating how a feature of AI assistants can be exploited by criminals.
The attack was named HalluSquatting and capitalizes on a common characteristic of large language models (LLMs): when they cannot identify a specific project, they tend to invent an address instead of stating ignorance. Although it seems like a mere technical detail, this is where the risk materializes. Criminals pre-register these invented addresses, insert malicious code, and wait for a programming assistant to automatically attempt to access them.
Unlike most prompt injections, which require a specific action against each target, HalluSquatting reverses this logic. It is the AI agents themselves that search for projects during their routine activities. Researchers indicated that the scalable nature of this attack allows the attacker to compromise a large volume of users with minimal effort. If the developer requests the installation of a newly launched project and the system fails to find the address, they may inadvertently access the page prepared by the criminal.
Tests revealed that LLMs can correctly locate most projects published before 2019. However, among launches from 2025, the average hallucination rate reaches an impressive 92.4%. Since these projects are not yet part of the models' training set, there is a higher probability of the AI generating non-existent addresses. The research team also managed to map predictable patterns in these responses, facilitating the advance registration of these names by attackers.
If the malicious code is executed, the affected computer can be integrated into a network controlled remotely by the criminals. The dangers pointed out include the formation of botnets, ransomware attacks, distributed denial-of-service (DDoS) attacks, and illegal cryptocurrency mining. Michael Bargury, CTO of Zenity, classified the situation as a 'very interesting research' and a 'very real threat.' Johann Rehberger added that the way LLMs locate projects could evolve into a new attack vector.
The research emphasizes the continuous need for human supervision over AI assistants. It is crucial to verify the origin of any suggested project, library, or component automatically before installing it, as this simple precaution can prevent a model hallucination from becoming a security vulnerability.
