ChatGPT’s Advanced Voice Mode launched in the UK this week - but not in the European Union. Although rumours about a "ban" have been swirling, it's believed that OpenAI may have held back its latest feature over fears that its emotion-sensing abilities could fall foul of the EU’s AI act, the world’s first major piece of AI legislation.
Advanced Voice Mode (which enables "live" conversations where the chatbot responds more like a human being) is able to read non-verbal cues such as the speed someone is talking at in order to respond with emotion. The EU’s AI Act prohibits "the use of AI systems to infer emotions of a natural person."
But how likely are such regulations to stifle innovation? And what is the "right" sort of regulation to enable enterprises to engage with AI? The Stack spoke to experts to find out.
It’s actually not clear that Advanced Voice Mode would be banned under the regulations, which would mean OpenAI is simply being cautious, says Curtis Wilson, staff data scientist at app security company Synopsys Software Integrity Group.
Wilson explains that similarly "careful" reactions were seen in the years after the General Data Protection Regulation (GDPR).
Wilson says: "It’s unclear if the EU AI Act actually prohibits Advanced Voice Mode at all. The part most people seem to point to is Article 5, particularly paragraph 1f which prohibits systems from inferring emotions. But this paragraph also specifies 'in the areas of workplace and education institutions' and the related recital makes it clear that the concern is poorly calibrated systems leading to discrimination against minority groups when the model misinterprets their emotions."
READ MORE: EU AI Act comes into force: Understanding the world's first artificial intelligence law
Companies will try to avoid being the "guinea pig" and falling foul of such regulations, and this could offer an opportunity for enterprises in the compliance sector, as more regulations come into force around the world, Wilson believes.
"One of the bigger directional changes I expect from the influx of global regulations over the next few years is to see a burgeoning AI regulatory compliance sector take shape to help companies manage a complex global AI regulatory environment."
Wilson believes the problem has been a lack of clarity, and this holds important lessons for future regulations.
He says: "Clarity is coming; Article 96 requires the Commission to provide guidelines on the practical implementation by August 2026 – 18 months after the rules on prohibited systems actually come into force. These guidelines should have been in place first.
"Developers need to know what is and is not covered by the regulation – preferably without having to employ the services of outside companies or legal firms. This is why clear, concise, and accurate guidelines (updated over time to match evolving technologies) are something I hope to see more of in the future.”
Compliance in the age of Generative AI
The case illustrates what will become one of the key challenges for global enterprises in the AI age, says Luke Dash, CEO at compliance company ISMS.online.
As more AI regulations come into force, enterprises will face issues if these regulations are not broadly similar across different territories.
Dash says: "Inconsistent regulations across regions will hinder AI deployment and complicate compliance for organisations operating outside of these areas. This will create fragmentation that forces companies to adopt region-specific strategies, potentially slowing global progress whilst also risking non-compliance and inconsistent implementation.
"Future regulations should focus on harmonising international standards to create a more coherent landscape."
While regulation is often seen as stifling growth, with AI, regulation will be essential to fostering acceptance of the technology, says Dr Kimberley Hardcastle, Assistant Professor Northumbria University.
As a result, regulation will be essential to integrating AI into the enterprise, and into society as a whole, she argues.
"Empirical studies, such as those conducted by the European Commission, indicate that well-structured regulations not only mitigate risks associated with bias and discrimination in AI but also enhance economic growth by creating a level playing field for innovation," Dr Hardcastle says. "Therefore, a robust regulatory landscape is not merely a hindrance but a facilitator that can drive sustainable and equitable AI adoption."
READ MORE: "Mediocre, but not completely incompetent": Top maths professor's ChatGPT o1 "Strawberry" review
Dr Hardcastle believes that AI may be so fast-moving it requires a new kind of regulation which can adapt to new challenges with "real-time adjustment".
Regulators must also learn the lessons of the social media age, she says.
She says: ‘The evolution of generative AI parallels the early development of the social media industry, where rapid innovation often outpaced regulatory responses, leading to significant societal repercussions.
"Likewise, the current landscape of generative AI reveals a competitive environment among companies to achieve artificial general intelligence at the expense of responsible development and ethical standards. This trajectory raises urgent concerns about potential harms, including biases in AI outputs and misuse of technology.
"To prevent history from repeating itself, we must learn from the social media experience, stakeholders must implement proactive regulatory frameworks that prioritise safety and ethics, ensuring that the pursuit of technological advancement does not compromise societal well-being."
