OpenAI, Google, Microsoft and Co. promise to introduce additional security measures for generative AI technology. One such measure is to use watermarks to indicate when something was created by AI.

The White House press service announced Friday night that it had reached an agreement with seven “leading AI companies.” They voluntarily promise, but perhaps under slight pressure from the government, additional measures to increase the security and reliability of their AI technology. Rules include OpenAI, Google and Microsoft and their popular AI tools.

Among other things, the companies promise to subject new systems to external security tests before they are made available to the general public. They must also actively investigate the potential risks of AI themselves and communicate with the government, but also with other industry players. In addition, sufficient investments must also be made in the security of the technology.

Trust in AI

A third part of the agreement is about strengthening trust between humans and artificial intelligence. Organizations will be open and transparent about what their tools can and cannot do, and provide guidelines for the safe use of the technology.

By adding watermarks to content, it aims to make it clearer for users whether something was created by or with artificial intelligence to stop the spread of misinformation and deepfakes. Finally, AI companies promise to only develop artificial intelligence for the greater good.

OpenAI, Google and Microsoft each say in their own statement that they fully support the proposed measures. Similar proposals are also on the table in Europe in the European Commission’s AI Act. However, this bill can expect less support in the business community for the time being. In fact, OpenAI Sam Altman has already warned that the regulations could make it impossible for his company to continue offering ChatGPT in Europe.