What is happening between the Pentagon, Anthropic, and OpenAI? Everything about the situation that favored the rise of Claude at the expense of ChatGPT

Recently OpenAI announced a collaboration with the United States Department of Defense – that is, the Pentagon – to provide artificial intelligence technologies to be used in activities related to national security. The news has reignited a debate that is already very strong in the United States: to what extent should artificial intelligence be used in the military sphere and in public security?

The issue returned to the spotlight when the Wall Street Journal revealed that Anthropic, the company that develops the chatbot Claude, had asked for guarantees about the use of its technologies by the Pentagon, with the aim of preventing its systems from being used in certain military operations without adequate limits. The clash with the Department of Defense arose precisely over these issues: the U.S. government was asking for greater freedom in the use of AI models, while Anthropic insisted on introducing stricter constraints. The disagreement between the two entities then led to the breakdown of their working relationship – with all the consequences involved (it is indeed likely that the matter will end up in court).

How OpenAI ended up working with the Pentagon

Collaboration between technology companies and the U.S. Department of Defense is nothing new: in recent years several Silicon Valley companies, including Google, Amazon, and Anthropic itself, have provided tools and infrastructure used in various operational areas of the Pentagon – currently led by the Trump-aligned Pete Hegseth. The tensions between Anthropic and the Department of Defense, however, have favored the rise of OpenAI, which has expanded its collaboration with the Trump administration. The move, however, has sparked numerous criticisms.

Among the most significant reactions is that of Caitlin Kalinowski, head of OpenAI’s robotics division, who announced her intention to leave the company. According to Kalinowski, Sam Altman’s recent decisions regarding collaboration with the defense sector would have required a deeper and less hasty evaluation. Despite OpenAI’s reassurances about the limits agreed upon with the Pentagon, there are no concrete guarantees that its artificial intelligence systems could not be used, for example, to facilitate forms of mass surveillance – an area that is poorly regulated in the United States, leaving room for various legal interpretations.

The consequences of the affair

The conflict between OpenAI and Anthropic also reflects two deeply different visions regarding the development of artificial intelligence. Dario Amodei, now CEO of Anthropic, previously worked at OpenAI, but in 2021 he left the company precisely to promote a more cautious development of artificial intelligence and one more attentive to security risks. In its early years OpenAI also supported a rather cautious approach to the development of AI. In recent years, however, the company’s strategy has changed: it has accelerated the release of new products and services, adopting a more aggressive and competition-oriented approach.

OpenAI is indeed looking with interest at the economic opportunities offered by a collaboration with the Pentagon, given that the artificial intelligence sector is not yet profitable enough to cover on its own the enormous costs required to develop and maintain these systems. The developments surrounding the Pentagon, however, have caused reputational damage to OpenAI. In addition to the resignation of one of its most important executives, the announcement of the agreement with the Department of Defense has led to a sharp increase in uninstallations of the ChatGPT app. In the same period Claude, developed precisely by the competitor Anthropic, became for the first time the most downloaded free app on the App Store in the United States.