Google sued by activist over AI-generated false claims

Conservative activist Robby Starbuck has filed a lawsuit against Google in Delaware state court, claiming the company’s AI systems generated and distributed false and defamatory statements about him to millions of users.

Starbuck alleges Google’s AI, including its Bard and Gemma chatbots, made serious accusations linking him to criminal activity and fabricated sources when responding to user queries.

Google spokesperson Jose Castaneda said most of the statements were due to “hallucinations” from Bard, a known issue with large language models, and the company worked to address the problem in 2023.

Starbuck is seeking at least US$15 million in damages.

He previously settled a similar lawsuit with Meta Platforms in August over AI-generated content.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

Section 230 may not shield AI-generated defamation

The Communications Decency Act Section 230 has shielded hosts of user posts, yet it may not cover AI firms when systems create harmful text 1.
Scholars say firms can face defamation liability if they knowingly or recklessly allow or ignore false statements presented as fact that harm a person’s reputation after notice 2.
OpenAI, maker of ChatGPT, has faced claims over ChatGPT hallucinations, which are confident but wrong answers; courts may view such firms as creators, not neutral hosts 1.
Starbuck v. Meta filed in April 2025 settled in August without a ruling; attention moved to his Google case where he seeks at least $15 million 2.

Enterprise AI buyers are prioritizing third-party guardrail solutions

Companies rolling out large language models face hallucinations and toxic replies, which drives demand for guardrails that screen prompts and outputs under rules like blocking defamation 3.
Future AGI Protect (AI safety tooling), Galileo AI (model evaluation and quality), plus Arize AI (model observability and monitoring) offer real-time spotting 4. They catch prompt attacks, hallucinations, plus policy breaches with monitoring dashboards 4.
Amazon Bedrock Guardrails sits inside Amazon Web Services’ Bedrock platform for foundation models. It adds grounding checks for hallucination detection; when paired with Bedrock Agents it can send low-confidence results to human reviewers 5.
Mindgard plus Netskope deliver automated red teaming for generative AI (GenAI) with data leak prevention to support compliance while stopping prompt attacks 6.
Machine learning operations (MLOps) tools with compliance features log every prompt and response to build audit trails for explainability that regulators expect 7.