Politico has published an extensive report claiming that, following Trump’s election, Apple changed its AI training guidelines on issues such as DEI, vaccines, elections, and Trump himself. Here are the details. Data annotation It is common practice for tech companies to rely on subcontractors to help with the labeling and post-training process of their AI models. Politico’s report says that Apple contracts Transperfect, a company that offers “translation services and solutions,” including AI data collection and annotation. Data annotation is one of the most important steps in AI training. In a nutshell, annotators label and categorize raw data, evaluate AI-generated outputs, and flag issues such as bias or safety risks. Their work produces structured feedback that, in turn, researchers can use to fine-tune or retrain models. According to the report, about 200 people work on data annotation for Apple’s AI models at Transperfect’s Barcelona office. Politico’s findings In its report, Politico says that it reviewed two sets of documents data labelers relied on for their work on Apple’s AI models: one that was used between 2024 and early 2025, and a second one, issued by Apple in March, with a set of new guidelines: “The March document introduced updated guidelines on how the AI talks about diversity, equity and inclusion policies — a topic for which boardrooms around the country, including Apple itself, have come under high-profile attacks from the U.S. administration — compared to the 2024 rating document seen by POLITICO. As well as taking greater care around a host of politically sensitive topics, including diversity initiatives, elections, and vaccines, the company also updated its guidance to give closer attention to the AI’s responses on Trump’s supporters and on Apple itself.“ Politico also says that sections on “intolerance” and “systemic racism” were removed. At the same time, topics such as DEI policies, Gaza, Crimea, Kashmir, Taiwan, vaccines, and elections were added to a list of “sensitive” topics, which already included LGBTQ+ rights, gun control, and general references to disputed territories. Finally, the report also says that the March document added more detail on how annotators should deal with Trump-related issues. However, there were no changes as to how the model should ultimately behave regarding him. Apple and Transperfect strongly deny the report, sort of In a statement to Politico, an Apple spokesperson denied that the company has changed its overall approach to AI policy, but didn’t deny the specific changes mentioned in the report: “Apple Intelligence is grounded in our Responsible AI principles, which guide every step from training to evaluation. Claims that we’ve shifted this approach or policy are completely false. (…) We train our own models and work with third-party vendors to evaluate them using structured topics, including sensitive ones, to ensure they handle a wide range of user queries responsibly. These topics are shaped by our principles and updated regularly to keep improving our models.” Apple also pointed to its Responsible AI Principles, a set of four guidelines that underpin its AI development, which can be found here. Likewise, Transperfect co-CEO and President Phil Shawe gave the following statement to Politico: “These claims are completely false, and we deny them in the strongest possible terms. (…) We regularly receive updated guidelines for our work, and over the last year, there have been more than 70 updates provided — none of these changed any policy, which has remained consistent.” As Politico noted, “(t)he statement did not specify which claims he was referring to.” What’s your take on how AI models should behave regarding increasingly polarized issues? Let us know in the comments. Accessory deals on Amazon