Ethics teams say they lack support as firms prioritize performance and product speed
Technology companies that publicly endorse the ethical development of artificial intelligence (AI) are not following through on those commitments, according to a new report by Stanford University researchers. Despite publishing guidelines and employing interdisciplinary teams focused on AI ethics, many firms continue to prioritize speed and performance over safety and responsibility.
The report, published by Stanford’s Institute for Human-Centered Artificial Intelligence, highlights that while companies often publicize their ethical intentions and employ researchers to develop technical and social safeguards, they rarely implement those measures in practice. Ethics staff frequently feel sidelined, especially when their work conflicts with business goals such as revenue generation or launch deadlines.
According to the report, many ethics professionals within tech firms describe an unsupportive or even hostile workplace culture. Product managers often view ethical interventions as harmful to business efficiency. One anonymous employee said, “Being vocal about slowing down AI development was risky—it simply wasn’t part of the process.”
The report did not name specific companies, but the findings come amid growing scrutiny of AI practices by governments, academics, and the public. Concerns around AI misuse—from data privacy breaches to algorithmic bias and intellectual property violations—have intensified since OpenAI’s release of ChatGPT and the subsequent rise of competing tools like Google’s Gemini.
Employees also noted that ethical concerns typically arise late in the product cycle, making it difficult to influence design decisions or correct issues. Constant team reshuffling further hinders long-term progress on ethical initiatives.
Stanford’s researchers pointed out that ethics recommendations often struggle to compete with performance benchmarks. “Recommendations that might lower model performance require irrefutable quantitative proof,” the report stated. “Yet ethical metrics are hard to quantify, and most firms lack the infrastructure to measure them reliably.”
