AI Roundup: March 30, 2026

1. Tennessee Woman Jailed Five Months After AI Facial Recognition Misidentification

Clearview AI’s facial recognition system matched Angela Lipps, a 50-year-old Tennessee grandmother, to bank fraud crimes committed in North Dakota — a state she had never visited. Fargo police arrested Lipps on July 14, 2025, and she spent over five months in jail before investigators confirmed she was buying cigarettes and depositing Social Security checks in Tennessee at the time of the crimes. The charges were dismissed on Christmas Eve, but Lipps lost her house, car, and dog during incarceration, prompting Fargo’s police chief to issue a public apology and ban the use of commercial facial recognition tools like Clearview AI. Source

2. AI-Driven Zero-Day Discovery Reaches Inflection Point

Multiple demonstrations and reports this week confirmed that AI systems can now autonomously discover novel software vulnerabilities faster than most human security researchers. AI security startup AISLE identified all 12 zero-day vulnerabilities in OpenSSL’s January 2026 patch, and XBOW’s autonomous platform discovered a critical RCE flaw in Microsoft’s Devices Pricing Program. Researchers warn that as these capabilities become more accessible, the window between vulnerability introduction and exploitation will continue to shrink, challenging traditional patch cycles. Source

3. Research Finds LLMs Elevate Expert Consensus and Moderate Polarization

A study published in the Financial Times found that large language models tend to “elevate expert consensus and moderate views” when users engage them on contentious topics, contrasting with earlier findings about AI sycophancy reinforcing existing beliefs. The research suggests that because models are trained on broad corpora that overrepresent mainstream expert positions, they may act as a moderating force in public discourse — though critics note this could also suppress legitimate minority viewpoints. Source

4. Iran’s AI-Generated Propaganda Videos Outperform White House Content

404 Media reported that Iran-linked accounts are producing AI-generated LEGO-style mockery videos that are outperforming official White House content in viral reach and audience engagement. The videos leverage generative AI to produce high-volume, culturally resonant propaganda at a fraction of traditional production costs, marking a shift in state-sponsored influence operations from text-based disinformation to visually compelling AI-generated media. Source

5. Autonomous Delivery Robot Crashes Chicago Bus Shelter

A Serve Robotics autonomous delivery robot crashed into and shattered a glass bus shelter in Chicago, sparking renewed local protests over sidewalk robot deployments. The incident follows growing tension between delivery robot operators and pedestrians in dense urban areas, with critics arguing that city sidewalks are not appropriate testing grounds for autonomous vehicles. The company stated it is investigating the incident. Source

6. Northwestern Researchers Build Autonomous Modular Robot Legs

Northwestern University researchers published a system for creating autonomous modular robot legs from single-degree-of-freedom links that can self-assemble into acrobatic walking machines. The system uses reinforcement learning to discover locomotion gaits for arbitrary link configurations, enabling rapid prototyping of legged robots without custom mechanical design — a step toward general-purpose robotic platforms that adapt their physical form to the task at hand. Source