Skip to content
Submit incident
AII-1002-49603
ConfirmedA court, a regulator or the organization itself has stated the core facts, or the record cites a court or regulator source.

Amazon Alexa gave pro-Harris responses due to missing safety override

Amazon's Alexa voice assistant reportedly gave detailed pro-Harris responses while refusing to provide reasons to vote for Trump, due to an oversight in safety controls rather than intentional programming. Amazon addressed the issue quickly after the viral video and acknowledged the failure to meet its own anti-bias standards.

3 September 2024 Β· Amazon Β· Discrimination and fairness
3 September 2024
AII-1002-496033 sources ↓
Quick facts
Category
Discrimination and fairness
Country
United States
Source
The Times of India
Date
3 September 2024
Drafted by
Raidu Incident Detection

What happened

In early September 2024, a viral video showed Amazon's Alexa providing detailed reasons to vote for Vice President Kamala Harris when asked, while refusing to provide similar responses about former President Donald Trump with the message 'I cannot provide content that promotes a specific political party or candidate.'

Amazon explained to House Judiciary Committee staffers that the system had manual overrides programmed for Biden and Trump, but not for Harris. From July 21 to September 2, users asked Alexa for reasons to vote for Trump 14,000 times compared to only 225 times for Harris, leading Amazon to judge the Harris prompt as low-priority.

Reported impact

The incident drew accusations of political bias against Amazon's voice assistant during the presidential election cycle. Trump campaign officials characterized it as 'big tech election interference.'

Index analysis

Amazon's approach of using manual overrides for high-frequency query topics created a gap in coverage when an unexpected candidate entered the race late in the election cycle. The system lacked programmatic safeguards to prevent biased outputs for prompts without explicit overrides.

Governance controls

Amazon apologized and stated it failed to meet its own standards for preventing political bias in Alexa. The company fixed the issue within two hours of the video going viral by implementing a manual override for Harris queries.

Amazon subsequently audited its systems to prevent Alexa from answering election-related questions and implemented manual overrides for all candidates and election-related prompts company-wide.

Classification

Category
Discrimination and fairness
Industry
Consumer products
Country
United States
Harm status
Potential harm
Intent
Unintentional
Timing
Post-deployment
Harmed
General public
Failure modes
Biased model, Agent misaction, Content moderation failure
Consequences
Public apology
AI systems
Amazon Alexa

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

The Times of India
Why was Amazon's Alexa giving pro-Kamala responses? Company informs house judiciary committee - Times of India
24 October 2024
Fox Business
Amazon Alexa's pro-Harris responses weren't pre-programmed: source
24 October 2024
Breitbart
Report: Amazon Source Claims Alexa Wasn't Programmed to Support Kamala Harris - Very Few People Asked
28 October 2024
Sep 2024Study finds LLMs make inconsistent, biased decisions on home surveillance videosConfirmedDiscrimination and fairnessMassachusetts Institute of TechnologyTech XploreMirage News+1Sep 2024Amazon Alexa provided biased election responses favoring Kamala HarrisConfirmedDiscrimination and fairnessLe Journal de MontrealPanAm PostAlerta Digital+3Sep 2026Amazon blocks Meta's Muse AI from making purchases without authorizationConfirmedSecurity and cyberattacksasinfobaeSep 2026Amazon Blocks Meta's Muse AI Shopping Agent Over Policy ViolationsConfirmedGovernance, accountability and labourPunto InformaticoJawaPos.comSep 2026US Senators Call for FTC Probe of Amazon and Walmart AI Chatbots Over Made in USA SuppressionAllegedGovernance, accountability and labourReutersThe Wall Street JournalQuartz+3