AI needs rules
EDITORIAL: The decision by OpenAI, Anthropic and Google to jointly develop an AI standards body is a telling acknowledgment of the risks accompanying a technology advancing at breakneck speed. That three of the companies driving this extraordinary race now see a need for common safety standards suggests that concerns once viewed as distant can no longer be easily dismissed. AI’s potential is vast, but so are the consequences of allowing its capabilities to outpace humans’ ability to manage them. The initiative comes amidst mounting warnings about AI risks. Former Anthropic researcher Jacob Coxon, who also worked at OpenAI, recently resigned after expressing fears that AI companies were “gambling with our lives”. Former Google DeepMind researcher Bilal Chughtai has warned that AI could potentially kill humanity, while Anthropic chief Dario Amodei has called for a slowdown in development. Whether such dire predictions come to fruition or not, they cannot simply be ignored when they are being raised by people who have worked deep inside the industry. Nor are the dangers confined to the apocalyptic possibility of AI becoming uncontrollable, destroying critical systems or being used to trigger nuclear catastrophes. The more immediate disruption is going to be economic and social. AI threatens to displace workers across an expanding range of occupations, alter the value of skills and concentrate wealth and technological power in a handful of companies. Copyright and intellectual property disputes are already raising fundamental questions about whether AI systems can be trained on creators’ work without consent or compensation. The enormous resource requirements of data centres are another emerging concern. Their rapidly expanding electricity consumption is placing pressure on power grids, while their need for cooling also requires huge amounts of water. Their concentration in particular regions can further strain local infrastructure and raise questions about who ultimately bears the environmental and economic costs of AI’s expansion. This makes the question of regulation unavoidable. However, an AI standards body convened principally by the companies developing the technology carries an inherent conflict. Industry expertise is indispensable to developing workable safety standards, but there is a risk that the interests of AI companies could take precedence over the broader public interest when those same companies have a hand in writing the rules. As chief of Cohere, a Canadian AI company, has argued, the issue is not whether AI needs guardrails, but who writes them and whose interests they protect. Governments therefore cannot abandon their regulatory responsibility here. Independent testing, transparency requirements, liability rules and safeguards against particularly dangerous applications must have public authority behind them. Regulation must also keep pace with a technology that does not respect national borders. US Senator Bernie Sanders has called for the US and China to sign a treaty, pausing the development of advanced AI systems. Whether such a proposal ultimately materialises is another matter, but the need for international cooperation is becoming increasingly urgent. No single country can establish effective global safeguards for a technology being developed and deployed worldwide. President Donald Trump’s dismissal of AI warnings as “hoaxes” is therefore a highly misguided response. One doesn’t have to subscribe to predictions of imminent human extinction to recognise that technological progress does not inevitably produce benign outcomes. AI could deliver extraordinary gains in medicine, scientific research, education and productivity. Precisely because its benefits could be so transformative, the dangers associated with it must be brought under control. The objective should not be to stop innovation, but to ensure that those driving the AI race do not become the sole arbiters of its rules. It is imperative then to ensure that its risks do not outrun humans’ ability to manage them. Copyright Business Recorder, 2026
EDITORIAL: The decision by OpenAI, Anthropic and Google to jointly develop an AI standards body is a telling acknowledgment of the risks accompanying a technology advancing at breakneck speed. That three of the companies driving this extraordinary race now see a need for common safety standards suggests that concerns once viewed as distant can no longer be easily dismissed.
AI’s potential is vast, but so are the consequences of allowing its capabilities to outpace humans’ ability to manage them.
The initiative comes amidst mounting warnings about AI risks. Former Anthropic researcher Jacob Coxon, who also worked at OpenAI, recently resigned after expressing fears that AI companies were “gambling with our lives”. Former Google DeepMind researcher Bilal Chughtai has warned that AI could potentially kill humanity, while Anthropic chief Dario Amodei has called for a slowdown in development. Whether such dire predictions come to fruition or not, they cannot simply be ignored when they are being raised by people who have worked deep inside the industry.
Nor are the dangers confined to the apocalyptic possibility of AI becoming uncontrollable, destroying critical systems or being used to trigger nuclear catastrophes. The more immediate disruption is going to be economic and social. AI threatens to displace workers across an expanding range of occupations, alter the value of skills and concentrate wealth and technological power in a handful of companies.
Copyright and intellectual property disputes are already raising fundamental questions about whether AI systems can be trained on creators’ work without consent or compensation.
The enormous resource requirements of data centres are another emerging concern. Their rapidly expanding electricity consumption is placing pressure on power grids, while their need for cooling also requires huge amounts of water. Their concentration in particular regions can further strain local infrastructure and raise questions about who ultimately bears the environmental and economic costs of AI’s expansion.
This makes the question of regulation unavoidable. However, an AI standards body convened principally by the companies developing the technology carries an inherent conflict.
Industry expertise is indispensable to developing workable safety standards, but there is a risk that the interests of AI companies could take precedence over the broader public interest when those same companies have a hand in writing the rules. As chief of Cohere, a Canadian AI company, has argued, the issue is not whether AI needs guardrails, but who writes them and whose interests they protect.
Governments therefore cannot abandon their regulatory responsibility here. Independent testing, transparency requirements, liability rules and safeguards against particularly dangerous applications must have public authority behind them.
Regulation must also keep pace with a technology that does not respect national borders. US Senator Bernie Sanders has called for the US and China to sign a treaty, pausing the development of advanced AI systems. Whether such a proposal ultimately materialises is another matter, but the need for international cooperation is becoming increasingly urgent. No single country can establish effective global safeguards for a technology being developed and deployed worldwide.
President Donald Trump’s dismissal of AI warnings as “hoaxes” is therefore a highly misguided response. One doesn’t have to subscribe to predictions of imminent human extinction to recognise that technological progress does not inevitably produce benign outcomes.
AI could deliver extraordinary gains in medicine, scientific research, education and productivity. Precisely because its benefits could be so transformative, the dangers associated with it must be brought under control.
The objective should not be to stop innovation, but to ensure that those driving the AI race do not become the sole arbiters of its rules. It is imperative then to ensure that its risks do not outrun humans’ ability to manage them.
Copyright Business Recorder, 2026