In one minute
- Almadía upheld David Silva Rojas’s literary prize on 6 November. Today, El Español publishes a test in which Pangram classifies a ChatGPT poem as mostly human. [1][2]
- DataVita says its centres receive all their power from the grid. China, meanwhile, has prohibited SoftBank’s acquisition of ABB’s Chinese robotics business. [3][8]
- Environmental organisations use AI and restrict some applications. Local opposition to data centres adds pressure to infrastructure development in the United States. [11][12]
Main story
CULTURE · AUTHORSHIP · FROM OCTOBER’S SUSPICION TO HUMAN REVIEW
Almadía upholds Silva Rojas’s prize after reviewing how he wrote his novel
On 23 October, El País reported that Pangram had classified 99% of Donde el agua trabaja, by Peruvian writer David Silva Rojas, as AI-generated. Almadía subsequently cancelled the Spanish book presentations. The outcome came on 6 November: according to El Cultural, the publisher concluded that the novel was not written with AI and upheld the decision of jurors Cristina Rivera Garza, Selva Almada and Mónica Ojeda. [1]
The review focused on material from the writing process; the author had said he kept a writing diary and different versions. Almadía explained that it applied human judgement and consulted technologists, lawyers and linguistic experts. It also encouraged El País to apologise. This is the publisher’s conclusion: the percentage accused; human evidence cleared the author before those responsible for retaining or withdrawing the prize. [1]
The new element in today’s El Español report is another contrast: Pangram rated a poem written with ChatGPT as mostly human. The program warned that the text was too short for a conclusion. That informal experiment illustrates a possible failure; it does not by itself evaluate the entire tool. [2]
Why it matters: classifying 99% of a text as artificial does not mean there is a 99% probability of fraud. Assessing authorship means considering the result alongside drafts, versions and accounts of the writing process. Almadía’s point was precisely that a detector alone is insufficient. If a number can cancel a book presentation, those using it must explain what it measures before turning it into an accusation. [1][2]
Other news
2. DataVita replaces its promise of direct renewable power with grid supply
SCOTLAND · ENERGY · REPORTED ON 7 NOVEMBER
DataVita’s project in Lanarkshire, near Glasgow, was announced with more than 1 GW of renewable infrastructure and direct private connections. Its current website says the centres will be fully supplied by the grid, leaving a direct connection to nearby renewables as a future possibility. The difference concerns the physical supply being promised, rather than simply the choice of words. [3][4]
According to The Guardian, DataVita points to a renewable PPA, a power purchase agreement, and denies misrepresenting its plans. Such a contract and a direct line from a renewable installation are different commitments. The public documentation allows the descriptions to be compared, although it does not establish exactly when the change was decided. [3][5]
The UK government’s announcement on 29 January cited £8.2 billion in private investment and more than 3,400 projected jobs, using figures attributed to DataVita. These include construction and operation: they are not all permanent positions already created. In August, the government also announced a £202 million National Wealth Fund guarantee, within a £300 million financing package. A guarantee supports financing; it does not amount to paying out a £202 million grant. [6][7]
For Spain and Norway, the useful question is who obtains grid capacity and on what terms. Reviewing a project means asking separately about the available connection, the generation timetable and supply contracts. The word “green” does not establish which infrastructure will be ready when the servers start running.
3. China prohibits SoftBank’s acquisition of ABB’s Chinese robotics business
ACQUISITIONS · INVESTMENT SECURITY · STATEMENT ON 6 NOVEMBER
SoftBank confirmed that on 4 November it received notice from China’s foreign investment security review body prohibiting the acquisition of ABB’s robotics business in China. The global purchase, agreed in October 2025, was valued at approximately $5.4 billion. That price covers the whole transaction, rather than only the Chinese business. [8][9]
One possible route forward is to separate the Chinese operations, according to Bloomberg, cited in a Reuters report on 7 November reproduced by MarketScreener. This is an option under consideration, not a completed agreement. SoftBank does confirm discussions with ABB about how to proceed. For those assessing technology acquisitions, the case shows that approvals also determine what can actually be bought: a global transaction may have to change shape to move forward. [8][10]
4. Environmental organisations use AI and cut processing tasks to contain its impact
CONSERVATION · PRACTICES REPORTED BY REUTERS ON 7 NOVEMBER
All fifteen organisations interviewed by Reuters said they used AI and were concerned about data centres’ environmental impact. In Kenya, Save the Elephants describes analysing aerial images in hours, compared with an earlier exercise involving around 200,000 images that took nine people nine months. The Jane Goodall Institute estimates that digitising chimpanzee notes in Tanzania will fall from 13–15 years to about two. Conservation International uses satellite data to alert community patrols to possible fires in Cambodia. [11]
There are also decisions about what to stop processing. The World Resources Institute limits generated images and videos and stores answers to avoid repeating identical queries. The Nature Conservancy processes images of fishing catches on board, reducing the videos it sends to data centres. These are results and practices described by the organisations to Reuters, not an audit of their overall environmental balance. [11]
For an organisation buying AI, the lesson is concrete: identify the task that warrants the computation, reuse results and choose where processing happens. These decisions can become part of the working procedure, with a named person responsible and an output that can be reviewed. Buying more capacity does not replace that choice.
5. Brief · Neighbours add pressure to US data centres
TechCrunch/Equity examines local opposition to these facilities after the elections. [12] The contrast with Scotland is clear: infrastructure needs acceptance as well as financing. The business lesson is to address permits, electricity and community participation from the outset; the journalistic analysis does not announce a national ban.
The signal in the noise
Trust requires verification. A percentage can trigger an investigation; a promise can justify considering an investment. What follows must be material, contracts and decisions that allow events to be reconstructed. In Silva Rojas’s case, listening to the author and examining his process changed the outcome. For infrastructure and conservation, the requirement is just as practical: what will be done, with which resources, and who will answer for it? For a legal team, asking for that information before deciding is a way of working, not a formality at the end.
Sources
[1] El Cultural · Almadía upholds the prize · 06-11-2026
[2] El Español · Pangram and the poem test · 08-11-2026
[3] DataVita · Lanarkshire AI Growth Zone
[4] DataVita · Original announcement · 29-01-2026
[5] The Guardian · DataVita and electricity supply · 07-11-2026
[6] GOV.UK · Announced investment and jobs · 29-01-2026
[7] GOV.UK · National Wealth Fund guarantee · 18-08-2026
[8] SoftBank · Decision on ABB’s Chinese business · 06-11-2026
[9] SoftBank · Global acquisition agreement · 08-10-2025
[10] Reuters / MarketScreener · Report attributed to Bloomberg · 07-11-2026
[11] Reuters / The Star · Environmental groups embrace AI · 07-11-2026
[12] TechCrunch / Equity · Local opposition to data centres · 07-11-2026