India’s Supreme Court Proposes Banning AI in Verdicts and Bail Decisions
India’s Supreme Court AI committee released draft rules that would completely prohibit AI from being used to decide verdicts, determine bail eligibility, or make sentencing recommendations — a clear boundary for AI in the judiciary.
The draft regulations, reported by the Indian Express and Economic Times, take a pragmatic approach: AI can be used in courts for document summarisation, legal research, translation, and case management — but decisions that affect a person’s liberty or legal outcome must remain the sole domain of human judges. The proposal also bans AI from weighing evidence or predicting case outcomes.
The committee cited concerns about algorithmic bias, lack of transparency in AI decision-making, and the fundamental principle that legal decisions must be explainable and appealable — characteristics current AI systems struggle with.
Why it matters: This is the most definitive judicial AI boundary set by any court system globally. It provides a model that other jurisdictions — including NZ — could adopt. The distinction between “AI as tool” (research, translation) and “AI as decision-maker” (verdicts, bail) is a framework any judicial system considering AI should study closely.
UC Canterbury Research Rethinks AI Literacy: Move Beyond “Learning to Use AI”
New research from the University of Canterbury is reframing how New Zealand schools should teach AI literacy — shifting the conversation from “learning to use AI tools” to “learning to think critically about AI across different ages and levels.”
The UC-led study argues that current AI literacy frameworks are too focused on operational skills (prompt engineering, tool selection) and neglect the critical thinking skills students need to evaluate AI outputs, understand limitations, and identify manipulation. The research was published on June 2 and is already influencing NZ curriculum discussions.
Key findings: younger students benefit from understanding AI as a tool that can make mistakes (not a magic oracle), while older students need frameworks for evaluating AI-generated information — source checking, bias detection, and understanding how training data shapes outputs.
Why it matters: This is the most practical AI education research to come out of a NZ university. The “learn to use AI” approach treats AI literacy like computer literacy in the 1990s — learning to click and type. The UC research pushes toward something deeper: learning to question AI, challenge AI, and know when not to trust AI. If this framework gets adopted in the NZ curriculum, it will shape how a generation relates to AI.
realestate.co.nz Launches AI Image Search for Homes — Visual AI Changes How Kiwis Find Property
realestate.co.nz launched an AI-powered image search feature on June 4, allowing users to upload a photo and find homes with similar design features, layouts, or architectural styles. It’s a practical application of computer vision that changes how property is discovered.
Instead of filtering by bedroom count or price range (the old search paradigm), users can upload a photo of a kitchen they like and see listings with similar countertops, cabinet styles, or layouts. The feature uses on-device processing where possible to protect user privacy.
The launch is powered by a custom-trained vision model, and realestate.co.nz says the response within the first 24 hours exceeded expectations.
Why it matters: This is a concrete example of AI changing everyday Kiwi behaviour — not through flashy agents or chatbots, but through a simple feature that makes property search more intuitive. The broader lesson: the most impactful AI applications aren’t always the most talked-about ones. Visual search is quietly becoming a default expectation across e-commerce, and NZ property is the latest domain.
Bupa NZ Invests in AI for Smarter, Safer Aged Care — Falls Detection and Behaviour Monitoring
Bupa New Zealand has deployed an AI platform across its aged care facilities, using computer vision and sensor data to detect falls, monitor resident movement patterns, and alert staff to potential issues before they become emergencies.
The system, announced on June 2, uses non-intrusive sensors (not cameras in private rooms) to learn residents’ typical movement patterns and flag anomalies — a resident who hasn’t moved from their chair in an unusual period, or a sudden change in walking speed that could indicate declining mobility.
Bupa NZ Clinical and Quality Director Sandy Turnwald said the platform helps staff focus their attention where it’s most needed, rather than relying on scheduled check-ins that might miss incidents between rounds.
Why it matters: Aged care faces a chronic workforce shortage in NZ. AI monitoring systems that augment human carers — rather than replacing them — are a pragmatic middle ground. The key ethical question (and Bupa seems to be navigating it carefully) is consent: residents must opt in, and the system is designed to flag potential issues, not surveil. This is AI in care done right — augmentation, not automation.
🔍 THE BOTTOM LINE
This week’s AI-Edu stories all share a theme: AI is entering institutions that have never dealt with it before — courts, schools, property searches, aged care — and each is grappling with the same fundamental question: where’s the line between AI as a helpful tool and AI as an inappropriate replacement for human judgment? India’s Supreme Court drew a bright line. UC Canterbury is asking schools to draw their own. Bupa and realestate.co.nz are demonstrating what “helpful tool” looks like in practice. The line is different everywhere, but the conversation is the same.
❓ Frequently Asked Questions
Q: Could NZ courts adopt India’s AI-in-court rules? Potentially. NZ has no equivalent regulation yet, but the Law Commission’s review of AI in the justice system is expected to address similar questions. India’s framework provides a precedent for the “tool vs decision-maker” distinction.
Q: How can NZ schools implement the UC Canterbury AI literacy framework? Start with age-appropriate conversations: primary students learn that AI can be wrong, intermediate students learn to source-check AI outputs, secondary students learn bias detection and training data analysis. Resources are being developed through the UC research group.
Q: Does realestate.co.nz’s AI image search share my uploaded photos? The company says on-device processing is used where possible, and uploaded images aren’t stored for model training without consent. Users should still be cautious about uploading images of interiors that might reveal personal information.
Q: What privacy protections does Bupa’s aged care AI have? The system uses non-camera sensors (motion detectors, pressure mats) rather than video in private rooms. Residents and families must consent to monitoring. The system learns normal patterns and flags deviations, rather than recording continuous video.
SOURCES
- Indian Express — Supreme Court AI draft rules ban AI in verdicts, bail
- Economic Times — SC’s draft rules allow AI in courts, bar it in decision-making
- University of Canterbury — New research rethinks AI literacy in education
- IT Brief NZ — realestate.co.nz launches AI image search for homes
- Business Scoop NZ — Bupa invests in AI for smarter, safer aged care
- RNZ — The real-world cost of AI
- NZ Herald — Tech Insider: Automated benefit decisions