Today’s five stories ask who controls knowledge, readers and enforcement as AI reaches lab-grown brain tissue, export rules, search, banking and the fight against scams.

1

Scientists use lab-grown brain tissue to process Braille signals. Could biological computing challenge AI?

Researchers at the University of Bristol have demonstrated that tiny brain organoids can process tactile signals that a computer can decode. The experiment opens possibilities for biological computing, although the technology remains at an early stage.

Researchers at the University of Bristol have demonstrated that lab-grown brain organoids can respond to electrical signals representing Braille letters, allowing a computer to identify the original information from their neural activity.

The research, announced on 16 October, was published in Frontiers in Artificial Intelligence.

The experiment used a tactile sensor to capture patterns associated with the 26 letters of the Braille alphabet.

Researchers converted those patterns into electrical signals and transmitted them to brain organoids, small collections of living neural tissue grown in laboratory conditions.

The organoids responded with patterns of electrical activity, which were subsequently analysed using a machine-learning classifier.

Individual organoids produced signals that could be decoded correctly approximately 61% of the time. Combining the activity of three organoids increased the accuracy to 83%.

The researchers also reported that the organoid-based system was more resistant to certain artificially introduced signal noise than a decoder operating directly on the obscured signals.

The results are detailed in the original peer-reviewed study, Encoding tactile stimuli for Braille recognition with organoids.

The organoids were less than one millimetre across and contained approximately 10,000 neurons alongside supporting cells.

They lack many of the structures found in a human brain and should not be understood as miniature functioning human brains.

The experiment demonstrates that biological neural tissue can receive and transform encoded information in a way that produces measurable output.

It does not demonstrate that the organoids understood Braille, learned to read or developed intelligence.

The researchers explicitly acknowledge that they have not established learning or long-term adaptation.

Their next objective is to investigate more complex tactile information and systems in which organoids receive feedback, allowing researchers to study whether biological networks can adapt over time.

The work belongs to the emerging field of biohybrid computing, which combines living biological material with electronic systems.

Researchers are interested in whether biological neural networks might eventually offer different approaches to information processing, potentially with lower energy requirements than some conventional computing methods.

However, the current experiment does not establish that organoid-based computing is more energy-efficient at a system level.

Maintaining living tissue, measuring its activity and processing its signals all require specialised equipment and resources.

The findings therefore represent an experimental demonstration rather than a practical alternative to conventional AI hardware.

Source: University of Bristol

2

China moves to restrict AI technology exports. Who controls the future of innovation?

Beijing has proposed new restrictions on the transfer of advanced technologies, including AI models, agents and computing infrastructure. The move adds another dimension to global competition over technological power.

China’s Ministry of Commerce and Ministry of Science and Technology have proposed changes to the country’s export-control framework covering 40 advanced technologies, including 23 related to artificial intelligence.

The official announcement, issued on 16 October, opens a public consultation on revisions to China’s catalogue of technologies prohibited or restricted from export.

The proposed additions span AI, biomedicine and photovoltaic technologies.

Within AI, they cover large models, autonomous agents, computing clusters and information-processing technologies.

According to the Ministry of Commerce’s accompanying explanation, the wider proposal identifies 14 technologies for export prohibition and 26 for export restriction across the three sectors.

The consultation remains open until 15 November.

While the catalogue revisions have not yet been finalised, China has introduced a temporary registration requirement for certain international technology transfers falling within the proposed categories.

The requirement covers transfers of technology ownership, overseas licensing and changes of ownership or effective control to foreign entities.

The ministries say AI technologies already in the public domain through open-source publication or academic research are outside the specified controls.

Beijing argues that the measures are intended to prevent the misuse or uncontrolled transfer of strategically important technologies while supporting legitimate international cooperation.

The restrictions come amid continuing competition between China and the United States over access to advanced AI chips, models and other technological capabilities.

American export controls have already limited China’s access to certain advanced semiconductors and manufacturing technologies.

China’s proposal would extend its own control over the international transfer of strategically important technical knowledge.

However, the two approaches should not be treated as identical. The Chinese measures concern particular categories of technology exports and ownership transfers, while American controls include restrictions on specific hardware and related capabilities.

The immediate implications for technology companies will depend on which proposals are adopted and how the final rules are applied.

Source: Ministry of Commerce of China

3

Google gives publishers more visibility in AI Search. But who controls the readers?

Google is making sources easier to find in its AI-generated answers. For publishers struggling to retain audiences, visibility is welcome, but it is not the same as control over their journalism.

Google has announced changes to AI Overviews and AI Mode, introducing clearer links to original sources, more prominent news displays and additional information about publishers.

In its 15 October announcement, the company said links embedded within AI-generated responses would appear more frequently and higher in the text.

Google is also testing dedicated sections for official websites and expanding news carousels, including articles from publications selected through its Preferred Sources feature.

The company reports that the number of people using its Preferred Sources feature has grown more than sevenfold since the beginning of the year.

For news organisations, these changes address an increasingly important problem. AI-generated search results can provide users with summaries of information without requiring them to visit the websites that produced it.

Google argues that its changes will make the web easier to explore. But the announcement provides no evidence that more visible links will translate into significant additional traffic for publishers.

That distinction matters for journalism, particularly independent publications that depend on readers reaching their websites.

Producing original reporting requires investment in journalists, research and verification. When AI systems summarise that work within their own platforms, the economic value may increasingly be captured by the companies controlling distribution.

The changes may improve the experience for readers who want to investigate sources. Whether they improve the financial position of news organisations remains an open question.

Source: Google

4

Brazil’s Itaú expands AI banking. Who benefits from the efficiency?

Brazil’s largest private bank wants AI to transform how millions of customers manage their money. But greater automation does not necessarily mean greater financial inclusion.

Brazilian bank Itaú is expanding ia.i, its conversational banking service, initially introduced to approximately 300,000 customers in July 2026.

A new OpenAI case study, published on 16 October, describes how the bank is developing the service using customer information, generative AI and an internal platform called IARA.

The bank wants to extend the service across its customer base by the end of 2026, allowing users to interact with banking products through text or voice.

The system can assist with accounts, credit, cards, insurance and other services, while connecting customers to human support when necessary.

Behind the interface is IARA, Itaú’s AI governance and orchestration platform, which manages access to different models according to internal policies.

OpenAI is among the providers whose models can be accessed through the platform.

Itaú also uses Customer DNA, a system that draws on financial information associated with approximately 70 million customers to provide more contextual assistance.

The bank says the approach is intended to simplify customers’ financial lives and make complex services easier to navigate.

Its AI strategy also extends to internal operations. The legal department uses a system working with more than 90 million documents and processing over two million tasks monthly.

According to the case study, Itaú reports efficiency improvements of between 20% and 80% in certain stages of those processes. These results have not been independently verified.

The technology offers opportunities to improve customer service and reduce administrative work. However, it also raises questions about how personal financial information is used and whether operational savings will benefit customers.

Source: OpenAI

5

Fifteen technology companies join crackdown on African scam networks. Who controls enforcement?

An international operation has led to arrests, disrupted millions of fraudulent accounts and rescued trafficking victims. The results show what cooperation can achieve, while raising questions about the power of private technology platforms in law enforcement.

Fifteen technology companies have joined law enforcement agencies from ten countries in an operation against online scam networks operating in West Africa.

The initiative, involving Meta, Coinbase, Google, Microsoft, OpenAI and other organisations, targeted criminal networks connected to investment fraud, impersonation, romance scams and other forms of online deception.

According to Meta’s announcement on 13 October, the operation resulted in the removal of more than 3.7 million Facebook and Instagram accounts, pages and groups associated with suspected scam activity.

Authorities also confirmed 28 arrests and the rescue of 117 people who had been trafficked and forced to work in scam operations in Ghana.

The operation was coordinated through the UK’s National Crime Agency, alongside American and international law enforcement partners.

Participating organisations shared information about fraudulent accounts, communications infrastructure and financial transactions to help identify criminal networks operating across multiple platforms and jurisdictions.

Meta reported sharing leads involving more than 30 suspected scam centres, while Coinbase said it had identified over 65 active victims and prevented more than $415,000 in potential financial losses.

The cooperation illustrates a practical difficulty in tackling online crime. Fraud networks often operate across national borders, using social platforms, digital payment services and other infrastructure controlled by different organisations.

No single company or authority necessarily possesses enough information to reconstruct the entire operation.

However, the figures should be interpreted carefully. The millions of accounts, pages and groups removed do not represent millions of individually verified criminals. Nor does their removal demonstrate that the underlying networks have been permanently dismantled.

Investigations and prosecutions remain the responsibility of public authorities.

Source: Meta