Anthropic and OpenAI reached an agreement with the US government to share artificial intelligence models

Two large companies active in the field of artificial intelligence have recently reached an agreement with one of the security agencies of the American government, according to which, in order to increase the security of their services, they will share data with the government of this country. Apparently, other companies, including Google, are negotiating with the aforementioned institution and will probably reach a similar agreement in the near future.

What is the purpose of sharing AI data with the US government?

The US Artificial Intelligence Security Institute is part of the US National Institute of Technology and Standards (NIST) whose mission is to create and publish rules and frameworks related to artificial intelligence systems. This institution is also responsible for testing and evaluating the potential harm of tools based on artificial intelligence and is continuously testing various services. The said institution was created in 2023 by the direct order of Joe Biden, and his deputy, Kamala Harris, pointed out in a speech its importance in detecting and dealing with cyber threats and misuse of artificial intelligence.

Sam Altman, the CEO of OpenAI, recently hinted at an agreement between his company and the US government to share data. Now it seems that the mentioned agreement has been finalized and Entropic has joined its competitor to improve its services in cooperation with the Artificial Intelligence Security Institute. The two mentioned companies provide their artificial intelligence models to the government so that the necessary evaluations can be done on them. Although the official news of the cooperation of other private companies in this field has not been published so far; However, in an interview with a Google spokesperson, Engadget reported that similar negotiations are underway with the government.

In her recent statement, Elizabeth Kelly, head of the US Artificial Intelligence Security Institute, emphasized the importance of safety in technological innovations and cited the institution’s collaboration with Anthropic and OpenAI for the advancement of artificial intelligence science. It is worth noting that the agreement of the two mentioned companies with the US government is more like a memorandum of understanding and will not result in a specific legal obligation for them. However, the government will be able to get early access to the original AI models from both companies before the public release. Also, the British Artificial Intelligence Security Institute will cooperate with the American government and important data will be provided to the European counterpart.

The safety and security of artificial intelligence services has become one of the important concerns of regulatory bodies, and new laws are continuously proposed and approved in this field. In this regard, the state of California has recently passed a law that mandates the review of safety standards for artificial intelligence services; Provided that the cost of providing processing power to set them up is more than 100 million dollars. Manufacturers have also been asked to include an emergency stop switch in their products so that access to them can be stopped immediately if needed. The law in question is mandatory for all technology companies, and the California prosecutor has threatened to sue and prosecute the violating companies.