IN THE AI OF THE BEHOLDER: WHERE DOES RESPONSIBILITY LIE?
25 Sep 2025 — Earlier in 2025, an Artificial Intelligence (AI)-powered chatbot
spouted fascist language and called itself ‘MechaHitler’ before user feedback led its developers to remove its posts and ban hate speech in its programming.
In this case of AI going rogue, what safeguards could have prevented this from happening? As we rely more and more on AI tools and technology, how do we assess and manage its potential harm?
"For the first time in human history, we are witnessing a technology capable of making autonomous decisions—sometimes without direct human oversight."
A Force for Good and Bad
These regulatory and ethical questions are what Raju Chellam, Fellow of NUS Computing Executive Education at the NUS SoC (School of Computing), addresses in a two-day in-person course for managers and executives.
Chellam is the Editor-in-Chief of the AI E&G BoK (AI Ethics & Governance Body of Knowledge) and Chair of Cloud & Data Standards at the IT Standards Committee1.
“For the first time in human history, we are witnessing a technology capable of making autonomous decisions—sometimes without direct human oversight,” Chellam says.
Identity theft and scams have benefited from AI's ability to process information at scale
The Black Box of AI
There have been numerous instances of AI being purposely used for harm. These include everything from deepfake images and identity theft to cyber scams, misinformation and phishing websites.
A major concern is the “black box” nature of AI: even developers often do not fully understand why AI systems behave the way they do.
For example, the case of the chatbot mentioned earlier raises pertinent questions:
- What safeguards could have prevented this?
- How do we assess harm when AI systems go rogue?
- Who is accountable when AI behaves in ways it was not designed to?
“There’s a growing concern about AI being misused—intentionally or unintentionally,” Chellam says. “Yet, at the highest levels of management, there is a general lack of awareness about the potential harm AI can cause. Decision-makers need to know the techniques and tools to help them to mitigate or take ownership when AI misbehaves.”
With the Asia-Pacific region poised to invest a staggering US$175 billion in AI and Generative AI by 2028 – up from US$37.5 billion in 2023 – the time may be ripe for guidelines and regulation to steer AI development towards minimising harm.
Governance Pillars for AI Development
To address these challenges, Chellam advocates robust governance and data provenance policies, including the following for organisations developing or deploying AI systems:
- Model AI Governance Framework: A practical guide for building trusted AI ecosystems, from IMDA (Infocomm Media Development Authority) website.
- AI E&G BoK: A collaborative initiative by the SCS (Singapore Computer Society) and IMDA. The various versions focus on Narrow AI, GenAI and comprehensive AI use.
- AI Verify: AI Verify Toolkit and Project Moonshot frameworks from IMDA, including a sandbox to evaluate AI applications safely and securely.
- Appointing a CAIEO: Role and responsibilities of a Chief AI Ethics Officer (CAIEO) to ensure accountability and provide a clear point of contact for AI-related concerns, feedback and resolution.
As AI becomes more embedded into our daily lives, our ethical and regulatory understanding must catch up too.
AI for the Public Good
Chellam expects AI to morph into more complex forms. Real-time AI embedded into devices and Artificial General Intelligence (AGI), where AI supposedly simulates human consciousness, are some upcoming developments.
As AI becomes more complicated and sophisticated, our ethical and regulatory understanding of AI’s effects on humans and businesses must evolve too.
“Only then can we ensure AI remains a tool in service of humanity, rather than a force that overrides it,” Chellam says.
Raju Chellam teaches the course ‘Role of Ethics in Artificial Intelligence’ at the NUS School of Computing. The course also covers AI frameworks and regulations in key markets worldwide, and is part of the Professional Certificate in AI and ML Foundation for Executives. The next run will be on 9-10 February 2026.
1 The ITSC is appointed by the Singapore Standards Council under the ambit of the national standardisation programme managed by Enterprise Singapore (ESG) and supported by the Infocomm Media Development Authority (IMDA).