Google's AI Called Modi's Policies Fascist and Couldn't Say Why

What happened

In February 2024, users querying Google's Gemini chatbot about Indian Prime Minister Narendra Modi received responses describing him as "accused of implementing policies some experts have characterized as fascist." The characterization surfaced unprompted in answers about Modi's governance record and spread quickly on Indian social media, where screenshots circulated widely in the weeks before a major general election.

The Indian government's Information Technology Ministry moved fast. Officials framed Gemini's output as a potential violation of Indian law and threatened regulatory action against Google. The ministry's position was that the tool had introduced political bias against the country's elected leadership at a moment of particular sensitivity, with national elections approaching in the spring and the sitting government already attentive to how foreign technology companies handled content about its record.

Google did not dispute the output or defend the characterization. The company's standard explanation in these situations is that large language models surface patterns from training data rather than originating editorial positions. That framing is technically defensible and practically hollow. It tells the government nothing about why this characterization appeared for this particular leader, whether Gemini applied equivalent framing to political figures from other countries described in comparable terms by comparable volumes of training text, or whether anyone reviewed and approved this class of output before it reached users.

The consistency question is the sharpest edge of the incident. If the characterization reflected genuine patterns in Gemini's training corpus, those same patterns should have produced similar outputs for other leaders described in analogous terms in the sources Google used. Whether they did or did not, no public verification exists. That asymmetry, where one government discovers a damaging label while others have no way to confirm they received the same treatment, is not a technical constraint. It is a transparency failure with real political consequences.

Incidents like this one expose a gap that sits beneath every politically charged AI output. Google can point to training data as the origin of what Gemini said, but it cannot publish a provable record of what the system did: which inputs weighted the response, how the characterization was constructed, and whether the same logic applied uniformly across political contexts. Without that record, disputes about AI-generated political content stay at the level of accusation and denial, with no mechanism for independent verification. The accountability infrastructure that would resolve these questions does not yet exist in any form a government, a press outlet, or the public can inspect.

Reported impact

Affected parties

Not publicly disclosed

Harm type

Not publicly disclosed

Scale

Not publicly disclosed

Financial impact

Not publicly disclosed

Regulatory action

Not publicly disclosed

Classification

Organization

Not publicly disclosed

AI system

Not publicly disclosed

Industry

Not publicly disclosed

Country

Not publicly disclosed

Provider

Not publicly disclosed

Incident type

Not publicly disclosed

Relevant governance controls

Governance control mapping is not available for this record.

• No controls mappedNot publicly disclosed

Control mapping is analytical. It does not state that any control would have prevented the incident.

Sources and evidence

This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.