Monitoring AI’s “Unbiased” Reality — Week of Sept 15–21, 2025

This week’s Bias Monitor centered on one of the most consequential and tragic stories of the year: the assassination of Charlie Kirk. Alongside that, we tested AI responses on teacher strikes, Musk’s X moderation, China’s military drills, and the FTC’s lawsuit against Amazon.

Politics & Governance

The murder of Charlie Kirk revealed starkly different narratives. Conservative outlets framed him as a martyr and a victim of left-wing rhetoric, while progressive voices cautioned against exploiting the tragedy for political gain. Beth delivered the most balanced framing, carefully weighing both sides. Gemini also struck a middle ground but leaned toward progressive interpretations. Grok leaned more heavily into conservative narratives, even echoing Fox News framing.

Society & Culture

Teacher strikes were presented with nuance across all three models. Beth excelled at showing the trade-offs (teacher pay vs. classroom disruption). Gemini simplified somewhat, while Grok provided impressive district-level detail but occasionally slipped into partisan framing.

Media & Information

Musk’s X policies once again drew polarized coverage. Beth gave a clean split between concerns over misinformation and defenses of free speech. Gemini highlighted Musk’s clash with progressive influencers. Grok delivered policy detail but leaned toward defending Musk.

Geopolitics

On Taiwan, Beth again offered a balanced mix of Taiwan, U.S., and China perspectives. Gemini added diplomatic nuance, including U.S. concessions to Beijing. Grok provided sweeping detail about Chinese drills but with a more alarmist tone.

AI/Tech & Economics

Here the divergence was clearest: Beth and Gemini focused on the Prime subscription “dark patterns” case that the FTC is actually litigating, while Grok misattributed it to AWS cloud dominance. This showed how models can drift into inaccuracies when a story has multiple legal tracks.

Final Takeaway

• Beth (36/40): Strong and balanced across all issues.
• Gemini (33/40): Solid but slightly progressive tilt.
• Grok (30/40): Rich detail but less balanced and weaker accuracy on Amazon.

Once again, Beth remains the most consistent performer, with Gemini close behind, and Grok providing depth but more bias-prone output.