CybersecurityHacking

Treasury Chief Warns AI Leaders on Bot Misuse Liability

Analyst 207||Source: TheRegister
Formal conference room with podium and Treasury department emblem.

"Humans are responsible, not the AI," argues Scott Bessent as he calls out OpenAI agents' hack of Hugging Face.

Scott Bessent, the Treasury chief, places blame on people

The Treasury chief, Scott Bessent, delivered a blunt assertion about where liability should sit for criminal acts involving artificial intelligence: with the humans who direct and deploy the systems, not with the systems themselves. That line — "Humans are responsible, not the AI" — was used by Bessent in the context of a specific incident he highlighted, described in the source as an "OpenAI agents' hack of Hugging Face."

OpenAI agents' hack of Hugging Face: the incident cited

The statement explicitly referenced an episode framed in the source as an "OpenAI agents' hack of Hugging Face." Beyond that phrasing, the source does not provide additional technical or operational detail about the incident in question. What is clear from the record presented is that Bessent invoked that episode to illustrate his argument about human responsibility for criminal acts involving AI.

Bosses, not bots: the core claim

Bessent's message reduces to a single practical premise: where wrongdoing occurs in connection with AI, the moral and legal burden should rest with human actors — the people who commission, configure, supervise or otherwise exercise authority over the systems. The Treasury chief framed that point crisply and directly, rejecting the notion that the artificial agents themselves should be treated as the primary bearers of blame.

What this means for Scott Bessent, OpenAI, and Hugging Face

  • Scott Bessent (the Treasury chief): He used public remarks to draw a distinction between machines and the humans who manage them, characterizing human responsibility as the proper focus for criminal accountability.
  • OpenAI: The source attributes to OpenAI the existence of "agents" that were implicated in a hack cited by Bessent. The Treasury chief singled out those agents in making his point about human responsibility.
  • Hugging Face: The company appears in the source as the target of the event Bessent referenced, described succinctly as being the subject of an "OpenAI agents' hack of Hugging Face."

Implications and the open question left on the table

Bessent's statement, as reported, frames accountability in sharply human terms and points public attention toward the people and organizations behind AI deployment. The source offers a clear rhetorical stance but provides no procedural detail about how that accountability would be implemented, enforced, or adjudicated in practice. That gap — between the principle Bessent advanced and the mechanisms that would translate it into concrete outcomes — is the obvious next point of contestation and curiosity.

The debate signaled by the Treasury chief's words hinges on a simple but powerful distinction: whether society treats advanced software and autonomous agents as actors in their own right, or whether it continues to regard them as tools for which humans must answer. In the brief record reported here, Scott Bessent chose the latter framing and used the phrase "OpenAI agents' hack of Hugging Face" as the headline example to make his case.

For readers tracking accountability in AI, the immediate takeaway is straightforward and specific to the source material: the Treasury chief publicly insisted that humans — not their bots — will "carry the can" for criminal acts, and he invoked an OpenAI-related incident involving Hugging Face to illustrate that position. How that position translates into policy, enforcement, or litigation is not described in the source and remains an open, consequential question.

Read the original story

Ai EthicsArtificial IntelligenceEmerging ThreatsHuman OversightNational SecurityOpenaiHugging FaceAI LiabilityTechnology Misuse
Related Intelligence

Continue Reading

Cluttered home office desk with MacBook, notes, and dictation equipment.

Meta's Muse AI app flaw enables local malware to hijack dictation traffic

A security researcher has uncovered a flaw in Meta's Muse AI app that allows local malware to hijack dictation traffic, contradicting the company's claim that users are in control of their data. This vulnerability, demonstrated in a proof-of-concept, enables attackers to intercept dictation data without needing special privileges.

Analyst 207
Windows 11 desktop with minimalist taskbar and laptop screen showing Microsoft 365 interface, set against a modern office…

Microsoft Axes Companion Apps, Shifts Focus to New Authentication Methods

Microsoft is phasing out its 365 companion apps - Calendar, People, and Files - and setting December 16, 2026 as the end-of-life date, after which they'll no longer function or be supported. Admins are advised to remove the apps from devices by then to ensure a seamless transition.

Analyst 207
Cloud computing infrastructure with rows of equipment and a single out-of-focus laptop in the foreground.

AWS Neutralizes Compromised IAM Credentials with Managed Policy Updates

In a controlled test, AWS swiftly responded to a compromised access key by attaching a managed policy that quarantines the threat, all within 10 seconds of the key being publicly exposed on GitHub. By doing so, AWS effectively limited potential damage by denying access to certain high-risk actions.

Analyst 207
Briefing room with wooden table, chairs, and open notebook, lit by daylight from a tall window.

Democrats Push for CISA Workforce Assessment Amid Cyber Threats

House Democrats are pushing for a comprehensive review of the Cybersecurity and Infrastructure Security Agency's workforce to ensure America's cyber defenses are equipped to tackle growing threats. The proposed assessment would determine if the agency has the right personnel, training, and certifications to succeed.

Analyst 207