TikTok-owner, ByteDance, says it has sacked an intern for “maliciously interfering” with the training of one of its artificial intelligence (AI) models.

But the firm rejected reports about the extent of the damage caused by the unnamed individual, saying they “contain some exaggerations and inaccuracies”.

BBC News has contacted ByteDance to request further details about the incident.

The Chinese technology giant’s Doubao ChatGPT-like generative AI model is the country’s most popular AI chatbot.

“The individual was an intern with the commercialisation technology team and has no experience with the AI Lab,” ByteDance said in a statement.

“Their social media profile and some media reports contain inaccuracies.”

Its commercial online operations, including its large language AI models, were unaffected by the intern’s actions, the company added.

ByteDance also denied reports that the incident caused more than $10m of damage by disrupting an AI training system made up of thousands of powerful graphics processing units (GPU).

Aside from firing the person in August, ByteDance said it had informed the intern’s university and industry bodies about the incident.

The social media giant has been investing heavily in AI technology, which it uses to power not only its Doubao chatbot but also many other applications, including a text-to-video tool called Jimeng.

I was delighted to catch sight of a headline that spoke of a “shot in the arm for active fund industry” (October 5). Could it be that active fund managers are finally showing that the application of highly rewarded brain power is paying off for those whose money they manage? Will the clients at long last have their yachts?

But no, plus ça change! It turns out after all that what the active fund industry is really, really good at is not the delivery of great value for its clients but relieving them of their wealth through extortionate fees. Bravo!

Andrew Mitchell
London W4, UK

Gillian Tett was, as ever, on the money when she wrote “Data centres alone won’t stop the AI energy squeeze” (Opinion, October 5). But beyond the need for joined-up thinking from the market and governments to increase energy supply, the article misses a faster and cheaper way to stop the energy squeeze caused by the growth in artificial intelligence — namely making data centres fundamentally more energy efficient.

Eric Schmidt, former Google CEO, noted that AI has an infinite appetite for energy, so increasing grid capacity is no doubt essential to realising AI’s full potential. However, AI moves faster than our ability to increase grid capacity. Increasing grid capacity requires enormous capital investment, governmental and regulatory change, and public approval — it is not a simple task, nor one achieved quickly.

Tett highlights the fact that market forces alone cannot solve this and notes the need for government to create connected grid capacity and adjudicate distribution of the limited energy supply fairly.

However, the article, and the wider debate, pays scant attention to the huge energy waste in data centres.

Take the network switch — the workhorse of every data centre, shuttling data packets between clusters of graphic processing units and central processing units. Network switches alone consume 20 per cent of a data centre’s total power requirement but state of the art technology now allows for that to be reduced to less than 1 per cent. Efficiency gains are waiting to be realised across every element of the data centre technology stack and should be prioritised.

Any increase in grid capacity must be twinned with making data centres more energy efficient and sustainable. Where technologies exist that use lower power and offer equal or better performance, these should be promoted and prioritised, by the industry — and yes, also by government through policy and regulation.

This can be achieved in part by earmarking a portion of the enormous capital slated to expand energy supply to support and incentivise the roll out of efficient data centre technologies. It can also be achieved by implementing regulation in the spirit of Germany’s recently passed Energy Efficiency Act which mandates power usage effectiveness levels for data centres, forcing owner and operators to build sustainably.

Market forces alone will not drive change — government support and incentives for data centre and AI companies to invest in technologies that enable energy-efficient, sustainable AI will be required.

Mark Rushworth
Chief Executive & Founder, Finchetto, Guildford, Surrey, UK

I much enjoyed the column by Catherine De Vries and her lamentation on the rise of political entrepreneurship as evidenced by Donald Trump and, presumably, Nigel Farage (“We are moving from democracy to ‘emocracy’”, Opinion, October 11).

It nicely balanced the more standard FT opinion page fare on that day — “How to rescue spinouts from the ‘valley of death’” and “Animal spirits of British business need to be lifted” — both of which argued that increased entrepreneurship is the key to salvaging the UK’s faltering economy.

However, it seems a little ironic that De Vries should hark back to a golden age when politicians supposedly focused on “facts and evidence” and not “rhetorical style” or “emotions and feelings”. The return to a better past is precisely the defining narrative of these very “political entrepreneurs”.

Make America Great Again, indeed.

Tim Gordon
Partner, Best Practice AI
CEO, Liberal Democrats 2011-17
London N1, UK

Your article “Ukraine considers lifting export ban on drones” (Report, October 10) highlights far-reaching issues surrounding the manufacture and supply of drones within Ukraine. It clearly shines a light on the failure of the Ukrainian government to support its drone companies through research and development and product development but the key issue of a lack of standardisation, while briefly mentioned, is fundamental to the success of any export plans.

While it is possible for Ukrainian drone companies to raise funds through exporting, this pathway will prove extremely limited if safety and quality standards for the manufacture of drones are not adhered to.

The drone industry has evolved at an exceptional pace, leaving regulators and standards-makers rushing to catch up. However, in recent years, safety and quality standards for the manufacture of drone technology across all domains (land, maritime and air) have been in development. One example of this is the International Standardisation Organisation (ISO) 21384 series of standards for the safety and quality of Uncrewed Air Systems (UAS) which were published in 2023.

While Ukraine is not currently a member of the ISO committee responsible for the development of these standards, they can obtain access to these standards and those developed by other standards bodies. The standards can be complex and confusing but, supported by its allies, navigating the options and developing a strategy for the Ukrainian drone industry is achievable.

Perhaps, rather than flooding the Ukrainian market with free European drone systems, reducing the need for in-country procurement, we should be guiding them in the art of defence drone procurement and the importance of standards. Rather than giving a man a fish . . . we should be teaching them to fish for themselves.

Robert Garbett
Chief Executive, Drone Major Group, London W1, UK