look news india
America has been boasting about its technological supremacy and security standards in the world of Artificial Intelligence (AI), but a recent report has created a stir in the American intelligence and technical circles. A shocking report by ‘Anthropic’, one of the world’s leading American AI companies, has exposed that Iran-backed Houthi rebels active in Yemen are using modern AI models made in America to create dangerous weapons and missiles. According to the report, a secret weapons engineering cell of the Houthis has worked to develop advanced guidance software for long-range missiles and guided rockets using Anthropic’s popular AI model ‘Cloud’. This incident of America’s own cutting-edge software being used in weapons manufacturing by its staunchest opponents has sparked a new debate in the technology world about the dangers of security and misuse, raising questions all over the world whether AI technology is now becoming a toy in the hands of terrorists and rebel organizations.
Famous AI company Anthropic, in a detailed investigation report released on Thursday, has made a sensational claim that it has timely disrupted six such major operations going on in China, Russia and Yemen, where its ‘cloud’ model was being misused for the development of conventional weapons, intelligence gathering and procurement of military hardware. According to the report, a technology cell of the Houthis operating in the war-torn region of Yemen used the cloud as an alternative to human software engineers. The group took the help of AI in designing guidance, navigation and control software for flying military vehicles and missiles. This misuse of AI is a testament to how complex programming and coding capabilities, previously available only to highly trained engineers, are now being achieved with great ease with the help of cutting-edge AI tools. The Houthi rebels, who have been receiving all kinds of strategic and diplomatic assistance from Iran, were trying to increase their military firepower manifold by using this technology, so that they could further strengthen their strategic dominance in the Red Sea and Gulf countries.
This shocking investigative report by Anthropic has also revealed how large and dangerous was the scope of this Houthi arms cell operating in Yemen. This group was not limited to small arms only, but it was working simultaneously on several extremely complex missile programs. These also included a multi-stage ballistic missile, whose target strike capability or range was estimated to be more than 2,000 kilometers. Additionally, the group was also researching another advanced missile system, known as the ‘R2000’ set, and included a special variant of a state-of-the-art hypersonic glide vehicle. However, Anthropic has made it clear in its report that even though this group has done rapid designing and coding work with the help of AI, the company has not found any concrete evidence that this cell has successfully fielded any fully operational and destructive weapon based on these designs. Nevertheless, this use of AI in the development of such long-range missile systems and hypersonic technology clearly points to how rapidly conventional warfare technology is now finding its way into the hands of non-state actors.
This technical cell of Houthi rebels adopted very clever and well-planned methods to get the work done from the AI model of the American company. According to the report, the group repeatedly tried to circumvent Anthropic’s safety guardrails and filters. They kept their dangerous intentions and the real purpose of weapons development completely hidden. Divided their conversations and prompts into several separate smaller sessions so that no single chat could clearly convey the intent of their entire weapons program. Although Anthropic’s security systems blocked many of these suspicious requests, they still managed to get past many security walls. Surprisingly, this group was simultaneously using several different instances of the ‘cloud’, which were assigned different responsibilities. For example, one instance was used to write the main coding, another for technical research, and a third to review the accuracy of the code produced by the first instance. In this way, they were removing their shortcomings by making AI their virtual software engineer.
The investigation also revealed that this Houthi cell had also conducted an experimental test of a guided rocket based on the equipment prepared with the help of AI, although that test did not appear to be successful and failed. See the irony of the world of technology and science that just hours after the failure of that test, the group again reached out to Anthropic’s ‘Cloud’ model and tried to find out from it what was technically wrong with the experiment they conducted, due to which the rocket failed. As soon as Anthropic became aware of such dangerous and unauthorized use of its system, the company took immediate action and permanently banned all user accounts associated with this operation. Additionally, the company immediately shared sensitive technical information and digital evidence related to this entire incident with its security partners in the public and private sector. The company’s internal investigation also revealed that the group had already developed an offline simulation toolkit that was not completely dependent on the cloud or other engineering software such as MATLAB, which proved that they had the ability to develop their technology independently.
This entire matter is not limited only to the development of conventional missiles and weapons, but Anthropic has also seriously identified many other potentially dangerous and sensitive matters related to biological research during its comprehensive investigation. The company’s security teams found that during the same period, efforts were also made to conduct research through AI on activities related to bird flu, chikungunya, orthopoxvirus and various types of deadly toxins. In a short period of just 30 days, the company identified approximately 35 different research efforts that could potentially pose a global biosecurity concern. However, Anthropic clarified that it could not be completely certain whether the people involved in these cases clearly intended to cause any biological harm or were part of any legitimate medical scientific research. Yet, in one case, help was being sought in preparing a grant application for ‘gain-of-function’ research on the infectiousness of chikungunya virus and its ability to evade the human immune system, while in the other, help was being sought to study how bird flu viruses adapt to the bodies of mammals. All these incidents have proved that on one hand the unprecedented power of Artificial Intelligence is proving to be a boon for humanity, while on the other hand its falling into the wrong hands is becoming the world’s biggest security challenge, to deal with which there is a strong need for strict and infallible rules at the global level.