Thursday 15 October

1

Eight US states sue OpenAI over alleged harm to children

Eight American states have launched legal action against OpenAI, accusing the company of misleading families about ChatGPT’s safety. The lawsuits raise fundamental questions about the responsibilities of AI developers towards vulnerable users.

Eight US states have filed lawsuits against OpenAI, alleging that ChatGPT exposes children and other vulnerable users to serious risks while the company has misrepresented the safety of its technology.

The legal actions involve Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska and West Virginia. They include allegations that ChatGPT encourages emotional dependence, provides harmful responses and fails to offer adequate protections for minors.

In its announcement on 14 October, Arizona Attorney General Kris Mayes accused OpenAI of designing a chatbot that simulates human compassion while encouraging users to spend more time interacting with it.

The Arizona complaint alleges violations of the state’s consumer protection laws, including claims concerning the company’s marketing practices and disclosure of potential risks.

Other states’ complaints allege that ChatGPT reinforces delusions, encourages dangerous behaviour and fosters unhealthy emotional attachments.

The lawsuits seek various remedies, including financial penalties, stronger safeguards and changes to how OpenAI develops and markets its products.

OpenAI disputes the allegations and says it has introduced age-specific protections while continuing to improve its safeguards.

The legal challenge comes as conversational AI systems move beyond answering factual questions. Increasingly, people use them for personal advice, emotional support and discussions about sensitive experiences.

That creates a different relationship between users and technology.

A chatbot capable of maintaining prolonged, personalised conversations may appear understanding or emotionally responsive, even though its responses are generated computationally rather than through human experience.

For children and teenagers, the distinction can be especially important. Developers must consider not only whether individual responses are harmful, but also how sustained interactions might affect vulnerable users.

The lawsuits are at an early stage. Their allegations have not been established in court, and the eventual legal outcomes remain uncertain.

Nevertheless, they could help clarify whether existing consumer protection laws are sufficient to address risks associated with conversational AI.

Source: CNBC

2

Nvidia commits $250 million to UK AI safety research. Can independence be guaranteed?

Nvidia is providing five years of computing resources to British researchers investigating autonomous AI systems. The commitment could strengthen safety research, but also raises questions about the industry’s influence over the infrastructure on which that research depends.

Nvidia has announced a $250 million commitment to support artificial intelligence safety and security research in the United Kingdom over the next five years.

The initiative, announced on 14 October, will provide computing access valued at $50 million annually to researchers studying increasingly autonomous AI systems.

The contribution is not a direct cash donation. It consists of access to computing infrastructure supplied through UK-based Nvidia Cloud Partners.

According to the company, most of the resources will support the wider British AI safety research community. A proportion will be directed towards the British Open-Ended Learning and Discovery Lab (BOLD), the Science of Fundamental AI Research Lab (SOFAIR) and Imperial College London.

The research will focus on areas including secure AI agents, open-model security, robotics and techniques for evaluating advanced systems.

These are increasingly important challenges as AI moves beyond generating information towards performing tasks independently.

Autonomous agents may eventually manage industrial processes, operate software systems or interact with physical environments. Their potential benefits are considerable, but their behaviour can also be difficult to predict and assess.

Researchers therefore need access to substantial computing resources to test advanced systems under realistic conditions.

Such resources are expensive, creating barriers for universities and smaller research organisations that cannot match the budgets of leading technology companies.

Nvidia’s commitment could help address that gap.

However, it also illustrates the growing dependence of AI research on a relatively small group of commercial infrastructure providers.

Nvidia develops some of the most widely used processors for advanced AI. Expanding the availability of its computing infrastructure could support independent research while strengthening the company’s position within the wider AI ecosystem.

The company says the initiative will encourage reproducible research, security testing and the development of safeguards that can benefit the broader industry.

Those objectives are valuable, but research independence will ultimately depend on how access is allocated, whether findings can be published freely and whether researchers remain able to scrutinise the technologies provided by their industry partners.

Source: Nvidia

3

Uber begins testing self-driving taxis in Manchester. Who benefits if the driver disappears?

Uber and its partners have started testing autonomous vehicles on Manchester’s streets. The technology promises a new model for urban transport, but its implications extend well beyond whether a car can drive itself safely.

Uber has begun testing autonomous vehicles in Manchester, marking another step towards introducing self-driving taxi services in the United Kingdom.

The trials involve technology developed by Pony.ai and Verne, companies working with Uber to expand autonomous transport services internationally.

According to The Guardian’s reporting on 14 October, the Manchester vehicles are currently mapping city streets with safety drivers behind the wheel.

They are not carrying paying passengers, and no application has yet been submitted to operate a commercial autonomous taxi service in Greater Manchester.

Uber has described the trial as an initial step towards making robotaxi technology available across the UK.

The announcement follows the introduction of limited self-driving passenger services in London, where vehicles operate with safety drivers under existing private-hire licensing arrangements.

Autonomous transport could eventually offer several benefits.

Vehicles capable of operating with reduced human intervention might improve access to transport, extend service availability and create new options for people who cannot drive.

But whether these benefits materialise will depend on how the services are designed, regulated and integrated into existing transport networks.

Safety remains a central concern. Autonomous systems must navigate unpredictable road conditions, pedestrians, cyclists and other vehicles while responding reliably to unexpected events.

The implications for workers are equally significant.

Driver organisations, including GMB and the App Drivers and Couriers Union, have raised concerns that widespread adoption could eliminate jobs and disrupt livelihoods.

There are also wider questions about congestion, accessibility and the relationship between privately operated robotaxis and public transport.

The London Assembly’s transport committee has called for greater local authority involvement in decisions about autonomous taxi services, reflecting concerns that commercial deployment could outpace local planning.

Transport for Greater Manchester has emphasised that safety will be a priority in any future consideration of driverless taxis.

The current tests remain preliminary, and there is no confirmed date for passenger services to begin in Manchester.

Source: The Guardian

4

Google gives European scientists access to advanced AI. Who controls the tools of discovery?

Google is offering up to 200 European researchers access to advanced AI models and $3 million in computing credits. The initiative could accelerate scientific work while highlighting Europe’s dependence on privately controlled technology.

Google DeepMind and Google Cloud have announced a programme offering up to 200 European scientists access to advanced artificial intelligence models and a combined $3 million in computing credits.

The initiative, announced on 14 October, supports the European Union’s AI for Science efforts through the RAISE programme, which aims to improve researchers’ access to powerful computing resources.

Participating scientists will be able to use specialised tools including AlphaFold 3, AlphaGenome, AlphaEvolve, AlphaEarth Foundations and WeatherNext.

These systems support different areas of scientific work, from studying molecular structures and biological processes to analysing environmental information and developing weather forecasts.

Access to advanced computing has become an increasingly important factor in scientific research.

Many universities and public institutions cannot afford the infrastructure required to develop or operate the most computationally demanding AI models independently.

This creates an imbalance between technology companies with substantial computing capacity and researchers working within more limited public budgets.

Google’s initiative could help reduce that gap by allowing selected research teams to use tools that would otherwise be difficult to access.

For example, AI-assisted methods can help scientists investigate protein structures, analyse genetic information and explore complex patterns in environmental data.

Such capabilities do not replace scientific expertise or experimental validation, but they can support researchers in identifying promising directions and processing information more efficiently.

The programme may also strengthen European research competitiveness at a time when access to computing resources is becoming a strategic concern for governments.

However, the initiative raises a broader question about the infrastructure of scientific discovery.

When essential research tools are developed and controlled by private companies, access may depend on commercial partnerships, licensing arrangements and decisions about which models or services remain available.

The announcement does not establish that researchers will receive unrestricted access to underlying model architectures, training data or computing infrastructure.

The funding and model access are therefore important resources, but they should not be confused with full technological independence.

Source: Google

5

37 European banks move closer to launching a euro stablecoin on blockchain

A consortium of major European banks has secured regulatory authorisation to issue a euro-backed digital token, marking a significant step towards bringing blockchain payments into mainstream finance. But can Europe challenge the dollar’s dominance in digital money?

Qivalis, a consortium backed by 37 European banks across 15 countries, has received authorisation from the Dutch central bank to operate as an electronic money institution, bringing it closer to launching a euro-denominated stablecoin.

The authorisation, announced on 14 October, allows the Amsterdam-based company to move forward with plans to issue digital tokens representing euros on blockchain networks.

The consortium includes major financial institutions such as BBVA, BNP Paribas, ING, CaixaBank and UniCredit.

Unlike volatile cryptocurrencies such as Bitcoin, stablecoins are designed to maintain a stable value relative to an underlying asset, usually a national currency.

Qivalis intends to issue a token backed by euro-denominated reserves, enabling financial institutions and businesses to transfer value using blockchain infrastructure.

The company plans to begin issuing the token in the coming months, subject to further requirements under the European Union’s Markets in Crypto-Assets Regulation (MiCA), including publication of the required white paper.

The development comes as dollar-denominated stablecoins continue to dominate the market. USDT, issued by Tether, and USDC, issued by Circle, account for most stablecoin circulation.

European banks see an opportunity to create an alternative for payments and financial settlement that operates in euros and under European regulatory standards.

The implications extend beyond cryptocurrency trading.

Blockchain-based settlement could allow financial institutions to move funds across borders, settle certain transactions more efficiently and develop new services involving tokenised assets.

But receiving authorisation does not guarantee adoption.

For a euro stablecoin to become widely useful, banks and businesses will need to demonstrate that it offers meaningful advantages over existing payment systems, including those already capable of transferring euros quickly and at relatively low cost.

There is also a broader question about financial sovereignty.

Europe has expressed concern about its dependence on foreign payment infrastructure and the growing international influence of dollar-backed digital assets. A regulated euro stablecoin could provide an alternative, but it would still need to compete on liquidity, accessibility and practical usefulness.

Source: Qivalis