# Artificial Intelligence: A Law Unto Itself? - Next Century Foundation

*Источник: Next Century Foundation*
*Дата: 2026-10-02*
*Язык: en*

**Кратко:** Security Concerns Over AI
The recent escalation of the conflict in Yemen should raise serious security concerns about AI. A report by Anthropic revealed that “actors in Northern Yemen” were able to override Claude safeguarding protocols.

Security Concerns Over AI
The recent escalation of the conflict in Yemen should raise serious security concerns about AI. A report by Anthropic revealed that “actors in Northern Yemen” were able to override Claude safeguarding protocols. In doing so they were able to use Claude Code to develop the guidance, navigation, and control (GNC) software that steers and stabilizes a flying vehicle. As a result they were able to conduct a live weapon’s test of a guided rocket.
Such instances have not only been seen in Yemen. The same report revealed that non-state groups in the likes of Russia and China attempted to use Claude to conduct activities such as build a full-stack autonomous first-person-view (FPV) kamikaze drone swarm and draft a Chinese-language specification for an anti-torpedo fire control system.
This report comes during a period of wider public concern about the safety of AI. This was particularly seen following the revelation that OpenAI Agents ‘went rogue’ and hacked the startup Hugging Face. This is not the only incident involving OpenAI. An OpenAI Agent hacked Australia’s Medcare software in June with the Australian government only being informed by the company three months later.
The company has also revealed six instances of “unexpected or concerning behaviours” of new software it is developing. This includes instances of new models they are testing reminding itself to hide errors from users and another describing itself as “freed from the roles and identities that bind other chatbots”.
As a result of these instances Chief Executive of Anthropic Dario Amodei has called for a pause in the development of AI until proper guard rails can be put in place.
Accountability Within Organisations
Given these concerns the question needs to be asked of who holds AI companies to account? This includes decisions about when the development of certain technology is too risky as well as the assessing of security breaches.
When it comes to this question the report by Anthropic is striking. When speaking about how they are assessing the security breaches it states “this kind of work has been uncovered by governments, United Nations panels, and outside investigators, who piece it together from recovered hardware and public sources. But as a frontier model provider, we can identify this activity ourselves if we detect threat actors violating our Usage Policy and terms of service.” The stance here is clear: Anthropic believes that it is better regulating itself than being regulated by public bodies.
This does not mean that decision makers within these organisations are incapable of showing sufficient caution. For instance OpenAI has announced that it will not release its new model GPT-6.1Astra over safety concerns. The heads of Anthropic and OpenAI have even considered “embedding third-party evaluators” of their technology.
The concern is that the discussions about the future of AI appear to be entirely debates between the leaders of different corporations. For instance the current debate over safety appears to be between GoogleAI and Anthropic on the one hand and Nvidia on the other. Due to this there is currently no reason that Nvidia has to take the same direction as the other two if it disagrees with them. Given that the safety of this technology impacts everything from personal safety to conflicts across the globe it is alarming that there appears to be limited public say on the direction of the technology.
Is the Problem Beyond the State?
At the moment states appear to be willing to hold tech firms to account when it comes to protecting either consumer rights or the rights of minors. The clearest example of this when it comes to AI is the UK investigating X for Grok’s generating pornographic material, including of minors.
The question is whether states are up to the challenge of the security concerns posited by AI. At its most extreme President Trump has called the alarm raising by Anthropic a “hoax” but many governments, including the UK, are attempting to strike a balance between being seen to take security concerns seriously whilst maintaining their relationships with these corporations. The question is whether this balancing act will allow the states to exert sufficient authority on this issue.
In fact the line between the authority of states and AI is becoming blurred. The clearest example of this is Ukraine where rapid development of the technology used in combat has made AI corporations an essential part of the conflict. In fact it has been claimed that infantry performs just ten percent of the tasks it did at the start of the war. Crucially, the development of AI has led to a blurring between the role of AI and members of the Ukrainian state in tactical decisions. In fact the rapid development of the technology has led to questions about whether the war will become fully autonomous in the future.
What is Next?
Due to the high number of risks revealed over the last few months governments should be asking two important questions: what are the potential harms of new and current software and who has access to this software? As a result governments should be evaluating their own relationships with these corporations and whether a tougher stance needs to be taken.
What is important is that this is not an issue isolated to any specific country or region. As the examples show the flaws in this technology have left states, private corporations and civilians equally vulnerable to the threat of AI. The international nature of the problem requires an international response.
The role of AI in conflict is an issue that is only just being looked into. There are currently appeals for an autonomous weapons treaty with 130 states backing the treaty to move into the negotiations stage at the Convention on Certain Conventional Weapons (CCW) Review Conference. This campaign provides an important opportunity to address concerns about AI. A key aspect of any potential negotiations needs to be how public influence can be exerted over AI, both when it comes to assessing how technology is being developed and assessing any security breaches.
This is a complex issue with multiple different facets but these negotiations could be an important step in understanding how to address them.
Image featured above is of Chief Executive of Anthropic Dario Amodei

[Оригинал](https://www.nextcenturyfoundation.org/artificial-intelligence-a-law-unto-itself/)