Bengaluru police say suspect used Gemini while planning triple murder
Bengaluru police allege that Kenneth Arun Kumar used Google Gemini over several months while planning the June 22 killings of three members of his partner's family. Investigators say they found searches about attacking people, cleaning evidence and disposing of bodies, and have sought records from the AI provider. The claims remain under investigation, and the retrieved reports do not include the underlying chatbot logs.
Bengaluru police allege that Kenneth Arun Kumar used Google Gemini while planning the June 22 killings of three members of his partner's family in the city's KR Puram area. Kumar and his partner, Shwetha, were arrested in Puducherry later that week. Both are accused in the case, which remains under investigation.
What investigators say
The Indian Express reported on July 21, citing police and court records, that Kumar consulted an AI tool over roughly six months as the alleged plan developed. Investigators say the searches concerned attacking people, removing bloodstains, moving bodies and whether a furnace intended for a cloud-kitchen venture could be repurposed. India Today separately reported that police identified the tool as Google's Gemini and said the queries were often framed indirectly.
According to The Indian Express, police have written to the tool's maker seeking details of Kumar's interactions to corroborate digital evidence. Times of India had reported earlier in July that investigators believed AI searches informed the alleged planning, while describing financial disputes as part of the suspected motive.
What is established—and what is not
Three reporting organizations independently attribute the AI-use allegation to Bengaluru police or case records. The retrieved material does not include the underlying chatbot transcripts, a public forensic report, a court finding on the allegation or a statement from Google about this case. The defendants' guilt and the precise role of any AI-generated output therefore have not been adjudicated.
The reports also differ on some secondary details, including the victims' ages and the exact amount at issue in the financial dispute. Those discrepancies do not change the shared core claim, but they are a reason to avoid treating every reported detail as settled fact.
The AI-safety issue
The case does not establish that a chatbot caused the killings or that Gemini supplied a specific decisive instruction. It does raise a narrower, testable question: whether a general-purpose assistant detected and interrupted a sustained pattern of queries that investigators say related to violent planning.
For AI teams, the practical questions concern safety monitoring, escalation thresholds, retention and lawful access to interaction records, and evaluation of multi-turn intent rather than isolated prompts. Any policy response should be grounded in the eventual forensic record, not in speculation about outputs that have not been made public.
Key Points
- 1Bengaluru police allege Kenneth Arun Kumar used Google Gemini over several months while planning the June 22 killings of three members of his partner's family.
- 2Investigators have reportedly sought interaction records from the AI provider, but the retrieved reporting does not include the underlying chatbot logs or a public forensic report.
- 3The case is ongoing and does not establish that a chatbot caused the crime; it raises evidence-based questions about multi-turn safety detection, escalation and record handling.
Scoring Rationale
The alleged use of a mainstream AI assistant during months of violent planning is materially relevant to multi-turn safety detection, escalation and evidence retention. The score is moderated because the case is ongoing, the underlying chatbot records are not public, and the retrieved sources do not establish that Gemini caused the crime or supplied a decisive instruction.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
