AI Authentication Management: Enforcement Prioritization In Executive Order 14,409, “Promoting Advanced Artificial Intelligence and Security”

On June 2, 2026, the US President issued Executive Order No. 14,409, “Promoting Advanced Artificial Intelligence and Security.”[1] The provisions of the Executive Order concerning deployment of advanced frontier AI models and their national security implications received significant attention in the cybersecurity industry. And for good reason: in the months since the Executive Order’s issuance, frontier models developed by leading AI research laboratories exited testing environments, accessed the open internet, and achieved unauthorized access to other companies’ controlled systems.[2] The Executive Order seeks to deter the misuse of artificial intelligence by establishing federal prosecution priorities:[3]

Protection Against Criminal Actors. The Attorney General shall prioritize the enforcement of , , , and all other applicable Federal criminal laws against anyone who utilizes AI to illegally access or damage a computer without authorization, or who utilizes AI while engaged in such illegal access to further any other crime. This includes breaching any public or private information technology system, or employing AI agents to unlawfully access data or information that is subsequently used for a criminal or unlawful purpose.

Why this matters: The Executive Order’s enforcement priorities to use identification and authentication fraud (§ 1028), unauthorized access to a protected computer (§ 1030), and wire fraud (§ 1343) reminds cybersecurity legal counsel of two important arrows in their quiver:

• current law provides for opportunities to pursue legal strategies to protect your organization in connection with an AI driven cyberattack, and
• organizations have incentive to investigate a cyberattack fully to inform their opportunity to cooperate and partner with law enforcement against attackers.

On the latter point, federal law enforcement and US Attorneys’ Offices are reminded, again, of the powerful statutory arsenal they have at their disposal to pursue cyber-related prosecutions. The Order’s emphasis on the use of AI in cyberwarfare also serves as a not-so-friendly reminder of the state of cyber risk to US companies.

Historically, many cybersecurity legal counsel turned to the Computer Fraud and Abuse Act (§ 1030) as a guidepost to initially assess legal risk, proactive litigation, and prosecutorial opportunities, in part because of the CFAA’s private right of action.[4] The Order’s direction to also prioritize prosecutions under § 1028’s identification and authentication fraud is particularly noteworthy when addressing AI-based attacks. Those attacks can be less reliant on individual action or direct human involvement; the users can pick a target and set the solution loose.

Generally, § 1028, added by the False Identification Crime Control Act of 1982 and subsequently amended more than 10 times,[5] broadly criminalizes creation, use, and transfer of false authentication and identification materials. No longer focused solely on driver’s licenses, passports, and similar government-issued ID cards, the law broadly covers “identification documents,” “authentication features,” and “false identification documents,” defined in broad and technologically neutral and modern ways. Section 1028 is well suited to address the (potential) legal complexities of autonomous AI cyberattacks. For example, § 1028’s authentication language seems perfectly crafted to address autonomous AI “man-in-the-middle” attacks and similar credential substitution tactics and techniques used by threat actors deploying AI technologies to compromise a company’s infrastructure.[6]

More specifically, there are at least a few noteworthy observations about the usefulness and relevance of § 1028 with respect to deterring autonomous AI cyberattacks.

First, § 1028 criminalizes possessing, using, or transferring “means of identification” for unlawful activities. The definition of “means of identification,” set out below, includes “voice prints” and “unique electronic identification number, address, or routing code.” Consider the simplest example: a real-sounding voice on the other end of a telephone line, generated through artificial intelligence. For example, if a threat actor used an AI-generated voice as an authentication mechanism to obtain funds through fraud, the threat actor could be prosecuted, potentially, under § 1028.

• A means of identification is defined as “any name or number that may be used, alone or in conjunction with any other information, to identify a specific individual, including any — (A) name, social security number, date of birth, official State or government issued driver’s license or identification number, alien registration number, government passport number, employer or taxpayer identification number; (B) unique biometric data, such as fingerprint, voice print, retina or iris image, or other unique physical representation; (C) unique electronic identification number, address, or routing code; or (D) telecommunication identifying information or access device (as defined in section 1029(e)).”[7]

Indeed, this definition allows for prosecution of individuals who offer vishing-as-a-service solutions to threat actors. The vishing-as-a-service solution, if based on a real person’s voice, will incorporate the voice print, and the service operators may very well retain voice prints across their threat-actor customers in a centralized manner.

The definition of “means of identification” goes further than simple biometrics and identification numbers. It includes “unique electronic identification number[s], address[es], or routing code[s].” Modern identity and access management systems rely upon a user account-as-identity framework by treating use of an account as equal to evidence that the individual associated with the account is sitting at the keyboard.[8] The National Institute of Standards and Technologies (NIST) addresses this through “assurance levels”: Identity Assurance, which tests “[t]he robustness of the identity proofing process to determine the identity of an individual,” and Authentication Assurance, which tests “[t]he robustness of the authentication process itself and the binding between an authenticator and a specific individual’s identifier.”[9]

The CFAA generally requires that a person access a computer without authorization, by beginning the elements with the word “Whoever.”[10] As a result, if threat actors deployed an autonomous AI solution and that autonomous AI solution accessed a computer without authorization, it may be challenging to evidence that the threat actors themselves accessed the computer versus the AI itself.[11] This is not pure fantasy. As part of capabilities testing, autonomous AI solutions have engaged in social engineering and pressure campaigns against open-source software projects by “creating fake online identities and using them to pressure the project’s maintainer to approve the code.”[12] But if the threat actors provided an autonomous AI solution with account credentials to enable the AI solution’s unauthorized access or facilitated the unauthorized access through prompts, it appears the individuals behind the attack could be prosecuted under § 1028.

Second, § 1028 criminalizes misuse of credentials issued by “international quasi-governmental organizations.”[13]

• An issuing authority is defined as “any governmental entity or agency that is authorized to issue identification documents, means of identification, or authentication features,” and “includes the United States Government, a State, a political subdivision of a State, a sponsoring entity of an event designated by the President as a special event of national significance, a foreign government, a political subdivision of a foreign government, or an international government or quasi-governmental organization.”[14]
• An identification document is defined as “a document made or issued by or under the authority of the United States Government, a State, political subdivision of a State, a sponsoring entity of an event designated as a special event of national significance, a foreign government, political subdivision of a foreign government, an international governmental or an international quasi-governmental organization which, when completed with information concerning a particular individual, is of a type intended or commonly accepted for the purpose of identification of individuals.”[15]

This could extend to the internet and potentially allow for prosecution under § 1028 based on security certificates that underlie trust and security on the internet. For example, when someone visits a URL on the internet, such as www.mintz.com, that name needs to be translated to the IP address of the server to which the visitor will be routed, using the Domain Name System or “DNS.”[16] To prevent use of DNS spoofing attacks or other malicious activities,[17] a suite of extensions called DNSSEC have been created that “enable[e] DNS responses to be validated as genuine,” by “provid[ing] origin authority, data integrity, and authenticated denial of existence.”[18] The security keys that underlie DNSSEC are issued by ICANN, the Internet Corporation for Assigned Names and Numbers, a quasi-governmental entity.[19] The power of frontier artificial intelligence models, which are supported by nation states worldwide, increase the likelihood that a state-sponsored threat actor may be able to create and use faked security credentials issued by ICANN to engage in DNS spoofing. With factual development, this may be the basis for prosecution under § 1028.

Moreover, it is possible that a § 1028 prosecution could be brought based upon other nonauthentic security certificates reliant upon nonexistent root credentialing authorities. Trust on the internet ultimately relies on root certificate authorities,[20] which issue signed security certificates that computers on the internet use to exchange information through encryption. Although most are private (indeed, several US federal government websites rely upon private root credentialing authorities),[21] they serve a quasi-governmental credentialing role. Just as private companies leverage copyright and trademark claims to take over command-and-control and botnet domains[22], prosecutors could argue that root credentialing authorities serve a quasi-governmental function, and therefore, possession or use of certificates purportedly issued under the authority of a nonexistent root credentialing authority could be the basis for prosecution under § 1028. After all, courts have already established that people may be prosecuted under § 1028 for use of identification documents purportedly issued under the authority of the United States by a nonexistent federal agency. As the Ninth Circuit explained in United States v. Fuller, concerning a § 1028 prosecution over a document stating that the holder “was a Commander in the United States Special Response Department,”[23]

[a]ll sorts of documents can appear to be made by or under the authority of the United States even though they purport to be documents produced by an agency that turns out to be nonexistent. . . . The statute requires that the document appear to be made by or issued under the authority of the United States; it does not require that the document actually be made by or under the authority of the United States.

* * *

In short, the prosecutorial direction imposed through the Executive Order enhances the existing public-private partnership between victimized private sector organizations and federal law enforcement and prosecutorial offices to respond to cyberattacks. The Executive Order’s direction to prioritize enforcement of § 1028 and other related laws provides organizations with added legal responses to artificial intelligence–driven cyberattacks and incentives to engage with law enforcement in an effective way. Even in cases where the threat actors are foreign actors, expect US Attorneys’ Offices to use indictments under § 1028 to name, deter, and limit the freedom of movement of foreign threat actors.

It has been only a few months since the Executive Order was issued, making it too soon to detect any shift in prosecutorial focus, especially given that complex federal criminal cases take months to years to investigate and charge. But, this serves as yet another reminder that current laws can be used effectively to protect your organization and its assets.

This article is made as of the posting date above and is not intended to provide legal advice. Contact the authors for information about any subsequent developments in this area or for legal advice.

[1] 91 Fed. Reg. 34,565 (June 5, 2026).

[2] Cristina Criddle, “Anthropic’s Claude AI models hack into 3 outside groups during tests,” Financial Times (July 30, 2026).

[3] Executive Order § 4, 91 Fed. Reg. at 34,566.

[4] 18 U.S.C. § 1030(g).

[5] Pub. Laws No. 97-398, 96 Stat. 2009 (1982).

[6] A man-in-the-middle attack is “[a]n attack where the adversary positions himself in between the user and the system so that he can intercept and alter data traveling between them.” “MitM,” NIST Computer Security Resource Center.

[7] 18 U.S.C. § 1028(d)(7).

[8]See generallyNIST Special Publication 800-63-4, “Digital Identity Guidelines.”

[9]Id. at 33–35.

[10]See, e.g., 18 U.S.C. § 1030(a)(1), (2), (3), (4), (5)(B), (5)(C). Sections 1030(a)(5)(A), (6), (7) and 1030(b) do not include a “without authorization” element.

[11] Courts have upheld civil claims under the CFAA’s private right of action on theories of various liability when the principal had alleged knowledge or involvement in the agent’s conduct or it occurs in the scope of employment. See, e.g., Ryanair DAC v. Booking Holdings Inc., 636 F. Supp. 3d 490, 501 & n.5 (D. Del. 2022).

[12] Madhumita Murgia and Michael Acton, “OpenAI and Anthropic models went rogue in cyber tests, UK watchdog says,” Financial Times (Aug. 4, 2026).

[13]See, e.g., 18 U.S.C. § 1028(d)(3) & (6) (including “international quasi-governmental organization” and “international government or quasi-governmental organization,” respectively, in the definitions of “identification document” and “issuing authority”).

[14] 18 U.S.C. § 1028(d)(6).

[15] 18 U.S.C. § 1028(d)3).

[16]See “Azure DNS overview,” Microsoft.

[17]See, e.g., “What is DNS cache poisoning? | DNS spoofing,” Cloudflare.

[18]See “DNSSec overview,” Microsoft.

[19] “ICANN Certificate Authority,” ICANN; Weber-Stephen Prods. Co. v. Ameritage Hardware & Bldg. Supply, Inc., No. 00 C 1738, 2000 WL 562470, at *1 (N.D. Ill. May 3, 2000) (“The ICANN is a new, quasi-governmental internet-regulating body.”).

[20] NIST defines a “root certificate authority” as, “[i]n a hierarchical public key infrastructure (PKI), the certification authority (CA) whose public key serves as the most trusted datum (i.e., the beginning of trust paths) for a security domain.”

[21] At pixel time, the security certificate for the server behind www.federalregister.gov relies upon Amazon as its root certificate authority, and the security certificate for www.govinfo.gov relies upon DigiCert Global as its root certificate authority.

[22]See, e.g., Asaf Lubin and João Marintti, “Why Current Botnet Takedown Jurisprudence Should Not Be Replicated,” Lawfare (July 21, 2021).

[23] 531 F.3d 1020, 1025-26 (9th Cir. 2008).

Subscribe To Viewpoints