With nearly 700 events happening daily, back room negotiations to monitor and dense texts, drafts and decision documents flooding emails at Cop31, an AI assistant almost sounds tempting.
The sales pitch for one app offered to journalists for free for the duration of the climate summit is certainly persuasive.
“It will turn press conferences, speeches and livestreams into clean transcripts with each speaker identified, and then into news drafts based only on what was said, in the tone you choose,” it promises.
“You will be able to proofread and translate into 83 languages within minutes.”
RM Block
Another app specialises in helping to get a foothold on the document mountain.
[ Mary Robinson warns countries at Cop31 against undermining IPCC scientistsOpens in new window ]
It offers to assess submissions by countries and parties, analyse their strength of language, identify “areas of alignment” and cluster them accordingly.
Quite apart from the danger of doing yourself out of a job, reliance on AI in this context carries the risk of having unknown app designers of unknown political views and agendas, potentially mischievous or malign, set the tone of news reports or nudge countries into alliances or opposing groups.
It’s not a risk any news organisation, analyst or official delegate worth their salt would take lightly.
But there is another problem with AI “helpers”: the carbon cost.
For every AI transaction, there’s a processor whirring in a data centre eating electricity most likely generated by fossil fuels and spewing out carbon emissions at the other end.

UN climate chief Simon Stiell addressed the issue head-on.
“AI’s benefits must outweigh its growing costs,” he told the summit. “The AI and wider tech industry must confront its growing energy and emissions footprint.
“What’s more, AI that spreads disinformation, undermines science, divides societies and concentrates wealth, damages the public trust on which the industry relies.”
So, AI is bad, right? Not if the programme of seminars, panel discussions and other side events at Cop31 is to be believed.
[ Fight to make climate polluters pay for crushing poor countries at Cop31Opens in new window ]
AI features mostly as a climate ally. It can aid early warning systems for communities vulnerable to extreme weather and other climate-related disasters.
It can pinpoint the best times to plant, water and spray crops, and balance electricity systems for maximum efficiencies.
A discussion on the climate impact of AI sought to find the truth, if indeed the truth can be found.
Ilias Iakovidis, European Commission adviser on the digital transformation, posed a question.
“If you introduce an IT solution for precision farming, what are you actually adding to emissions through putting sensors and drones and all types of gadgets that have negative impact on climate?”

Naturally, there’s an app for that.
Bob Burgoyne is from Carbon Trust, a company that deploys AI to compare the carbon emissions saved by a solution against the emissions incurred by that solution.
“If you invent a widget that saves some energy, perhaps through data-driven optimisation, what the methodology does is answer the question, are the emissions associated with that saving more than those incurred through making the widget itself but also any other emissions that might arise through its introduction?” he said.
[ Thousands protest at alternative ‘people’s Cop’ march in TurkeyOpens in new window ]
He gave the example of a project undertaken by a Swedish housing authority that wanted to install temperature sensors in 21,000 apartments to better match heating with the needs of the occupants, with flexibility that would prevent waste.
That’s a lot of digital information to be processed around the clock but Carbon Trust’s AI tools crunched the numbers and proved net emissions savings.
The case is not always so clear-cut, Burgoyne admitted. Not all AI companies are transparent about the energy they use so the carbon cost is hard to calculate.
Enter the Antalya Artificial Intelligence Pledge. Drawn up by the Turkish hosts of Cop31, it asks countries, cities, businesses, banks, investors, research institutions and international agencies to voluntarily take a stand on AI.
It’s a largely supportive stand, that recognises AI “already supports mitigation, adaptation and resilience and can accelerate climate policy implementation and scale up solutions”.
But it also notes “AI’s growing demand for electricity, computing infrastructure, water, land, and critical materials”.
Signatories declare “shared ambition to make AI part of the climate solution – sustainably powered, efficiently designed, transparently measured and available to all”.
Ireland has not signed the pledge “yet”, said the Department of Climate on Tuesday.
Beyond Fossil Fuels, an alliance of 60 NGOs across Europe, was not impressed with it.
“It brings no evidence that AI can benefit climate action,” the group said.
“A climate pledge that ignores what is needed to fix the footprint of AI’s infrastructure risks becoming a green label for business as usual.”
While the AI evangelists are becoming cosy at Cop31, a little exercise shows the technology may not be Holy Grail of digital fixes they claim it to be.
When the first draft of the summit’s cover decision – or final agreement – was published on Monday afternoon, its rambling mishmash of viewpoints and commitments left many observers perplexed.
Put through AI detectors, one declared it 87 per cent AI written, one found it 23 per cent AI and two said it was penned by humans.
The carbon cost of this inconclusiveness is not known.















