ai

Microsoft signs multi-billion-dollar AI pact with Nebius

Shares in Amsterdam-based Nebius surged more than 50% in US premarket trading after the company announced a $19.4 billion deal with Microsoft. The agreement will provide cloud computing power for AI workloads and is set to run until 2031.

Under the pact’s terms, Nebius will generate $17.4 billion in revenue over the next six years, with Microsoft also retaining the option to purchase additional computing capacity.

Investors responded strongly, with Nebius shares climbing 60% in extended trading on Monday and continuing their rise on Tuesday. Rival AI infrastructure company CoreWeave also benefited, gaining more than 6% premarket.

Nebius, spun out from Russian internet firm Yandex in 2023, specialises in supplying GPUs for training AI models. The deal underscores the high demand for computing power as companies race to develop increasingly advanced AI systems.

Industry leader Nvidia recently reported soaring earnings and forecast that AI infrastructure spending could reach $4 trillion by 2030.

The boom in AI infrastructure comes amid mounting questions over whether valuations are sustainable. OpenAI has reportedly reached a $500 billion valuation, while rival Anthropic recently raised $13 billion at a $183 billion valuation.

Some analysts, including OpenAI chief executive Sam Altman, have warned that the AI sector may already show signs of a bubble.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

AI adoption drops at large US companies for the first time since 2023

Despite the hype surrounding AI, new data suggests corporate adoption of AI is slowing.

A biweekly survey by the US Census Bureau found AI use among firms with over 250 employees dropped from nearly 14 percent in mid-June to under 12 percent in August, marking the largest decline since the survey began in November 2023.

Smaller companies with fewer than four workers saw a slight increase, but mid-sized businesses largely reported flat or falling AI adoption. The findings are worrying for tech investors and CEOs, who have invested heavily in enterprise AI in the hope of boosting productivity and revenue across industries.

So far, up to 95 per cent of companies using AI have not generated new income from the technology.

The decline comes amid underwhelming performance from high-profile AI releases. OpenAI’s GPT-5, expected to revolutionise enterprise AI, underperformed in benchmark tests, while some companies are rehiring human staff after previously reducing headcount based on AI promises.

Analysts warn that AI innovations may have plateaued, leaving enterprise adoption struggling to justify prior investments.

Unless enterprise AI starts delivering measurable results, corporate usage could continue to decline, signalling a potential slowdown in the broader AI-driven growth many had anticipated.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Anthropic AI faces legal setback in authors’ piracy lawsuit

A federal judge has rejected the $1.5 billion settlement Anthropic agreed to in a piracy lawsuit filed by authors.

Judge William Alsup expressed concerns that the deal was ‘nowhere close to complete’ and could be forced on writers without proper input.

The lawsuit involves around 500,000 authors whose works were allegedly used without permission to train Anthropic’s large language models. The proposed settlement would have granted $3,000 per work, a sum far exceeding previous copyright recoveries.

However, the judge criticised the lack of clarity regarding the list of works, authors, notification process, and claim forms.

Alsup instructed the lawyers to provide clear notice to class members and allow them to opt in or out. He also emphasised that Anthropic must be shielded from future claims on the same issue. The court set deadlines for a final list of works by September 15 and approval of all related documents by October 10.

The ruling highlights ongoing legal challenges for AI companies using copyrighted material for training large language models instead of relying solely on licensed or public-domain data.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Orson Welles lost film reconstructed with AI

More than 80 years after Orson Welles’ The Magnificent Ambersons was cut and lost, AI is being used to restore 43 missing minutes of the film.

Amazon-backed Showrunner, led by Edward Saatchi, is experimenting with AI technology to rebuild the destroyed sequences as part of a broader push to reimagine how Hollywood might use AI in storytelling.

The project is not intended for commercial release, since Showrunner has not secured rights from Warner Bros. or Concord, but instead aims to explore what could have been the director’s original vision.

The initiative marks a shift in the role of AI in filmmaking. Rather than serving only as a tool for effects, dubbing or storyboarding, it is being positioned as a foundation for long-form narrative creation.

Showrunner is developing AI models capable of sustaining complex plots, with the goal of eventually generating entire films. Saatchi envisions the platform as a type of ‘Netflix of AI,’ where audiences might one day interact with intellectual property and generate their own stories.

To reconstruct The Magnificent Ambersons, the company is combining traditional techniques with AI tools. New sequences will be shot with actors, while AI will be used for face and pose transfer to replicate the original cast.

Thousands of archival set photographs are being used to digitally recreate the film’s environments.

Filmmaker Brian Rose, who has rebuilt 30,000 missing frames over five years, has reconstructed set movements and timing to match the lost scenes, while VFX expert Tom Clive will assist in refining the likenesses of the original actors.

A project that underlines both the creative possibilities and ethical tensions surrounding AI in cinema. While the reconstructed footage will not be commercially exploited, it raises questions about the use of copyrighted material in training AI and the risk of replacing human creators.

For many, however, the experiment offers a glimpse of what Welles’ ambitious work might have looked like had it survived intact.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

OpenAI study links AI hallucinations to flawed testing incentives

OpenAI researchers say large language models continue to hallucinate because current evaluation methods encourage them to guess rather than admit uncertainty.

Hallucinations, defined as confident but false statements, persist despite advances in models such as GPT-5. Low-frequency facts, like specific dates or names, are particularly vulnerable.

The study argues that while pretraining predicts the next word without true or false labels, the real problem lies in accuracy-based testing. Evaluations that reward lucky guesses discourage models from saying ‘I don’t know’.

Researchers suggest penalising confident errors more heavily than uncertainty, and awarding partial credit when AI models acknowledge limits in knowledge. They argue that only by reforming evaluation methods can hallucinations be meaningfully reduced.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

Mistral AI pushes growth with new funding and global deals

Founded in 2023 by ex-Google DeepMind and Meta researchers, Mistral has quickly gained global attention with its open-source models and consumer app, which hit one million downloads within two weeks of launch.

Mistral AI is now seeking fresh funding at a reported $14 billion valuation, more than double its worth just a year ago. Its investors include Microsoft, Nvidia, Cisco, and Bpifrance, and it has signed partnerships with AFP, Stellantis, Orange, and France’s army.

Its growing suite of models spans large language, audio, coding, and reasoning systems, while its enterprise tools integrate with services such as Asana and Google Drive. French president Emmanuel Macron has openly endorsed the firm, framing it as a strategic alternative to US dominance in AI.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

Mental health concerns over chatbots fuel AI regulation calls

The impact of AI chatbots on mental health is emerging as a serious concern, with experts warning that such cases highlight the risks of more advanced systems.

Nate Soares, president of the US-based Machine Intelligence Research Institute, pointed to the tragic case of teenager Adam Raine, who took his own life after months of conversations with ChatGPT, as a warning signal for future dangers.

Soares, a former Google and Microsoft engineer, said that while companies design AI chatbots to be helpful and safe, they can produce unintended and harmful behaviour.

He warned that the same unpredictability could escalate if AI develops into artificial super-intelligence, systems capable of surpassing humans in all intellectual tasks. His new book with Eliezer Yudkowsky, If Anyone Builds It, Everyone Dies, argues that unchecked advances could lead to catastrophic outcomes.

He suggested that governments adopt a multilateral approach, similar to nuclear non-proliferation treaties, to halt a race towards super-intelligence.

Meanwhile, leading voices in AI remain divided. Meta’s chief AI scientist, Yann LeCun, has dismissed claims of an existential threat, insisting AI could instead benefit humanity.

The debate comes as OpenAI faces legal action from Raine’s family and introduces new safeguards for under-18s.

Psychotherapists and researchers also warn of the dangers of vulnerable people turning to chatbots instead of professional care, with early evidence suggesting AI tools may amplify delusional thoughts in those at risk.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Anthropic settles $1.5 billion copyright case with authors

The AI startup, Anthropic, has agreed to pay $1.5 billion to settle a copyright lawsuit accusing the company of using pirated books to train its Claude AI chatbot.

The proposed deal, one of the largest of its kind, comes after a group of authors claimed the startup deliberately downloaded unlicensed copies of around 500,000 works.

According to reports, Anthropic will pay about $3,000 per book and add interest while agreeing to destroy datasets containing the material. A California judge will review the settlement terms on 8 September before finalising them.

Lawyers for the plaintiffs described the outcome as a landmark, warning that using pirated websites for AI training is unlawful.

The case reflects mounting legal pressure on the AI industry, with companies such as OpenAI and Microsoft also facing copyright disputes. The settlement followed a June ruling in which a judge said using the books to train Claude was ‘transformative’ and qualified as fair use.

Anthropic said the deal resolves legacy claims while affirming its commitment to safe AI development.

Despite the legal challenges, Anthropic continues to grow rapidly. Earlier in August, the company secured $13 billion in funding for a valuation of $183 billion, underlining its rise as one of the fastest-growing players in the global technology sector.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Mistral secures €1.3B ASML investment amid $14B valuation

ASML has reportedly become the top shareholder in French AI company Mistral after investing €1.3 billion. The deal forms part of a wider €2 billion funding round that values Mistral at $14 billion, marking a significant milestone for the Paris-based startup.

The Dutch chip-making equipment giant will also gain a board seat at Mistral, with Bank of America advising on the investment. The move is seen as a step towards reinforcing European technological sovereignty by reducing reliance on American and Chinese AI systems.

The partnership could help Mistral expand its generative AI tools and open-source platforms while enhancing ASML’s ability to integrate data analytics into its operations.

Industry analysts suggest the collaboration will unite two European technology leaders at a critical moment in the global race for AI dominance.

Founded by Timothée Lacroix, Guillaume Lample, and Arthur Mensch, Mistral has quickly become one of Europe’s most valuable AI startups.

The company, backed by investors including Microsoft, Databricks, and General Catalyst, develops open-source generative AI models that directly compete with those produced by OpenAI.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Google avoids breakup as court ruling fuels AI Mode expansion

A US district judge has declined to order a breakup of Google, softening the blow of a 2024 ruling that found the company had illegally monopolised online search.

The decision means Google can press ahead with its shift from a search engine into an answer engine, powered by generative AI.

Google’s AI Mode replaces traditional blue links with direct responses to queries, echoing the style of ChatGPT. While the feature is optional for now, it could become the default.

That alarms publishers, who depend on search traffic for advertising revenue. Studies suggest chatbots reduce referral clicks by more than 90 percent, leaving many sites at risk of collapse.

Google is also experimenting with inserting ads into AI Mode, though it remains unclear how much revenue will flow to content creators. Websites can block their data from being scraped, but doing so would also remove them from Google search entirely.

Despite these concerns, Google argues that competition from ChatGPT, Perplexity, and other AI tools shows that new rivals are reshaping the search landscape.

The judge even cited the emergence of generative AI as a factor that altered the case against Google, underlining how the rise of AI has become central to the future of the internet.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!