Anthropic revealed China's AI military engineering project: anti-flag fire control, Taiwan air defence targets and targeted energy information

Chinese military researchers have used the United States artificial intelligence model as an aid to weapons development. Anthropic released threat intelligence report in September this year, with at least several Chinese-related actors using Claude to assist in the preparation of anti-flag fire control programmes and research into electricity @Ambassah: #Feb14: #SirFirthFeb14: #Feb14 #Feb14 #Feb14 On 11 October, the Lutherans Dr. Xian's Voice commented on the renewed focus on the weaponization of AI and biotechnological misuse, noting that advanced models are becoming new tools for national security competition.

Anti-flag fire control: from software to military-industrial procurement programme

Anthropic identified one of these China-related activities as GTG-17001. The actor used Claude to advance three lines of work for the anti-fowl system: to prepare specifications for fire control in Chinese, to develop more than 200 pages of technical proposals and to collate public information on the United States Navy anti-smuggling and anti-fowl project. The company assessed that the actor might be associated with Chinese military-industrial manufacturers who wished to provide equipment programmes to the Chinese Navy. AI is not here to produce weapons in place of the factory, but rather to accelerate the formation of technical documents, software components and procurement programmes.

Another activity, GTTG-17002, is more directly related to military risks in the Taiwan Strait. Anthropic disclosed that a Chinese researcher associated with China used a model to assist in the creation of analogue software for electronic warfare and anti-aircraft suppression systems, and that the project had set Taiwan ' s military objectives as a test scenario. The company assesses the actors ' links to the relevant systems of Chinese research institutions, including the Military Academy of Sciences of the Liberation Army, based on accounts and related information. These are the attribution assessments of the Platform survey and do not imply that the relevant software has become deployable weapons.

Targeted energy weapon intelligence: another path to technology acquisition

Activities No. GTG-17003 are not directly manufactured weapons. The report describes a Chinese associate who requested Claude to gather public information about high-powered microwave weapons and their supply chain abroad and to write an intelligence bulletin in Chinese for internal circulation to the Communist Party, the military or the national security leadership. The actor tried to identify suppliers of critical equipment, track business ownership relationships and compare foreign systems with existing projects in China.

Three cases point to the same change: the sorting of information, technical documentation, model comparison and programme review that in the past required a multi-person division of labour, can now be repeated in the same AI service. (a) Beijing to promote long-term civil-military integration and to closely link civilian research and military requirements; When researchers are able to access advanced models from abroad, this movement of technological capabilities no longer depends solely on hardware exports and laboratory cooperation.

Bigger alert for biological research and sanctions circumvention

The Anthropic report also revealed cases of misuse by pathogens, sanctions circumvention and other State actors, including high-risk research applications involving the Kikunkenya virus, and the use of models through intermediary platforms. The participants in these cases are not the same and cannot be attributed to Beijing for all abuses; They illustrate, however, that sensitive knowledge and software capabilities may cross traditional control boundaries through network services.

Lutherans has looked at the report under the CCP Global Security Strategy, and the most valuable part is not to mix all risks into a conspiracy, but to identify specific technology acquisition mechanisms. China's associated actors have left traces of the use of anti-flag, Taiwan's air defence and targeted energy intelligence in platform records. Model developers have blocked accounts, but there is no guarantee that actors will not switch to other models or local systems. For the Chinese military-industrial system, AI lowered the threshold for some of the technical jobs; This makes it more difficult to control the border between the integration of Chinese civil-military and technical services outside the territory of the Communist Party in the context of international security.