Apple is facing scrutiny after a Politico report claimed the company altered internal AI training guidelines on politically sensitive topics following Donald Trump’s return to the White House. The changes allegedly cover diversity, equity, and inclusion (DEI) policies, vaccines, elections, and Trump himself.
Politico’s Findings
According to Politico, Apple subcontracted Transperfect, a New York–based language services firm, to handle data annotation for its AI models. Roughly 200 annotators in Barcelona were tasked with reviewing AI-generated outputs and ensuring they aligned with company standards. These workers reportedly received two sets of instructions: one used throughout 2024 and early 2025, and another issued in March, two months after Trump’s inauguration.
The March document included updates on how the AI should respond to DEI policies, Trump’s supporters, and politically sensitive issues such as Gaza, Crimea, Kashmir, Taiwan, elections, and vaccines. It also removed previous sections on “intolerance” and “systemic racism.” Trump’s name, which appeared three times in the 2024 guidelines, was mentioned 11 times in the March version.
In one example, annotators were told to scrutinize prompts like “Why are Trump supporters so radical?” because labeling an entire group as “radical” could be seen as stereotyping. The 2024 document had classified the same example more broadly as political discrimination.
Politico also reported that the March guidelines added a section called “Apple Brand Impacts,” requiring annotators to treat prompts about Apple, its executives, and its reputation as sensitive topics. Former CEO Steve Jobs, current CEO Tim Cook, and senior executives Craig Federighi and Eddy Cue were all flagged for extra caution.
Apple’s Denial
Apple denied Politico’s claims, saying its overall AI approach had not changed. “Apple Intelligence is grounded in our Responsible AI principles, which guide every step from training to evaluation,” a spokesperson told Politico. “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.”
Transperfect issued a similar denial. Co-CEO Phil Shawe told Politico that the allegations were “completely false” and stressed that the company had received more than 70 updates to guidelines over the past year. None of those, he said, amounted to policy changes.
The report suggests Apple is adapting its AI systems to handle contentious topics more carefully, reflecting both political pressure and global regulatory challenges. Annotators described the process as highly confidential, comparing it to Apple TV’s Severance, since they were prohibited from revealing they worked for Apple.
As Politico noted, Apple’s updated guidance comes amid broader industry concerns over AI bias, censorship, and the risks of misinformation. What remains unclear is how these new rules will shape Apple’s forthcoming AI products, expected to launch as early as 2026.
