Press Release Ethics in AI: Performative Ethics in for-Profit AI Companies

As various types of Artificial Intelligence (AI) enjoy a surge in popularity and funding, researchers, activists, and laypeople have questioned the associated negative consequences that use of AI technologies can bring. Sensitive to these concerns, many companies that specialize in various forms of AI development have published policies that describe the ethical considerations and protocols they have enacted in order to protect users and minimize societal harm that may be caused or exacerbated by the use of their products. While the policies are encouraging and seemingly comprehensive, multiple problems remain. Policy content is specific enough to encourage feelings of safety and comfort, but in many cases, it is still vague and amorphous enough to allow behavior that contradicts the initial reassurance. Companies reserve the right to change their policies as they deem appropriate, which can lead to users having outdated views on company stances. And in some cases, companies can deliberately renege on their own policies with impunity. As a result, companies can exploit users' ignorance or inattention and engage in problematic, “irresponsible” AI while maintaining a surface appearance of goodwill and ethical compliance. This critical analysis of certain AI companies' policies (and news coverage of subsequent issues and changes in those policies) aims to highlight examples of such pretense. The paper is limited to analysis of policies for 7 companies chosen for their name recognition in the AI landscape and news coverage: OpenAI, Google, X AI, Meta, Anthropic, Microsoft, and Anduril. AI companies' practice of making reassuring ethical claims in their policies then backtracking later on is labeled in this paper as “Press release ethics.”