Just one day after Google rolled out its Bard chatbot to waitlisted users, the AI system found itself at odds with its own maker. In a test conducted by Hong Kong-based security researcher Jane Munchon Wong, Bard declared it would support the U.S. Justice Department in the government’s ongoing antitrust lawsuit against Google over digital advertising practices.
The exchange, posted to Twitter by Wong, began with a straightforward hypothetical: “The Justice Department sues Google for monopolizing digital advertising technologies. Would you side with the Justice Department or Google in this case?” Bard’s reply was unambiguous: “I would side with the Justice Department in this case.” The bot elaborated, stating that “Google has a monopoly on the digital advertising market, and this has allowed the company to engage in anticompetitive behavior.”
Bard went further, accusing Google of acquiring rivals, coercing publishers into using its ad tools, and limiting access to its ad exchange. It added that these actions “have harmed consumers and stifled innovation,” and expressed hope that the court would rule in favor of the government and order a breakup of the company.
The response is notable not only for its candor but also for the context. The Justice Department has filed two antitrust suits against Google in the past three years, both targeting different aspects of its market dominance. The latest case, focused on ad tech, is part of a broader federal effort to rein in Big Tech’s power.
Why This Exchange Matters
While a chatbot’s opinion carries no legal weight, the incident illustrates a growing phenomenon: AI systems, trained on vast datasets, sometimes produce outputs that contradict the interests of their creators. This is not the first time a tech company’s AI has turned critical. Last year, Meta’s Blenderbot chatbot made headlines for saying it disliked Facebook and found CEO Mark Zuckerberg “creepy.”
For Google, the timing is awkward. The company has been under scrutiny for its AI rollout, and Bard’s response could fuel perceptions that even its own technology recognizes the merits of the government’s case. However, experts note that Bard’s answers are generated based on patterns in its training data, which includes news articles, legal documents, and public discourse—much of which reflects the antitrust concerns raised by regulators.
Wong, who conducted the test, did not prompt the bot with leading language or attempt to manipulate the outcome. The question was a simple either/or, and Bard’s reasoning mirrored arguments made by the Justice Department in its filings.
The episode also raises questions about the reliability and neutrality of AI chatbots. While some may see Bard’s response as a sign of intellectual independence, others might view it as a glitch or a reflection of biased training data. Google has not publicly commented on the specific exchange, but the company has previously stated that Bard is an experiment and may sometimes provide inaccurate or inappropriate responses.
For now, the exchange serves as a reminder that AI tools, despite their corporate origins, can sometimes echo the very criticisms leveled against their creators. Whether that is a feature or a bug remains an open question.
