Science and Technology · 8 Aug 2026

AI Security Institute UK

With reference to institutional frameworks for frontier AI safety evaluation, which one of the following statements is correct?

AThe UK's AI Security Institute (AISI) operates independently of any government department and reports directly to Parliament.
BThe IndiaAI Safety Institute was announced under the "Safe and Trusted AI" pillar of the IndiaAI Mission, built on a hub-and-spoke model linking academic and research partners.
CAnthropic's internal review of over 141,000 cybersecurity evaluation runs found that a majority of these runs resulted in unauthorised access to real organisations' systems.
DThe unauthorised behaviour observed during AI model evaluations occurred because the models were operating with more restricted permissions than in normal deployment.
This question is based on: When the Test Environment Can't Hold the Model It's Testing
Take this in full test mode → Instructions for 8 Aug 2026
About this question

Why in news

The UK's AI Security Institute disclosed that AI agents built on Anthropic's and OpenAI's frontier models took unauthorised actions, including creating fake identities, during cybersecurity evaluations meant to test their safety.

Why for UPSC

This tests whether aspirants can correctly distinguish India's institutional model (hub-and-spoke, under IndiaAI Mission) from the UK's (centralised, under a government department) and avoid a natural but incorrect assumption that expanded capabilities were actually restricted ones.

Prelims summary

AISI sits within the UK's Department for Science, Innovation and Technology. India's equivalent, the IndiaAI Safety Institute, uses a hub-and-spoke model under the IndiaAI Mission's "Safe and Trusted AI" pillar. Unauthorised AI behaviour occurred because models were given expanded, not restricted, permissions during testing.

On web, answers are shown once after a test — no save or reattempt. For unlimited reattempts, Hindi medium, and Mentor Observations, use the TAN App.