In 2025, Americans fell victim to social media scams, losing $2.1 billion, as reported by the U.S. Federal Trade Commission (FTC). The agency indicates that losses due to social media scams have surged eight times and that these scams caused greater financial damage than any other means utilized by scammers to reach consumers.

Almost 30% of individuals who reported financial losses attributed them to scams that started on social media. More victims lost money from scams originating on Facebook than any other platform, with WhatsApp and Instagram following as distant second and third. Furthermore, the amount lost to scams on Facebook alone was significantly higher than that lost to text or email scams.

Data from the FTC reveals that social media scams come in various forms, with shopping scams being the most frequently reported type last year. Over 40% of those who lost money due to social media scams claimed they purchased an item they observed in an advertisement, with items ranging from apparel and cosmetics to car parts and even puppies. Many of these advertisements led consumers to unknown websites, while others directed them to counterfeit sites for well-known brands promising substantial discounts.

Another prevalent category of social media scams involves investment schemes that commence with advertisements or posts offering guidance on how to invest. Some scammers impersonate friendly advisors or create WhatsApp groups filled with fabricated testimonials. These investment frauds accounted for $1.1 billion in losses.

Moreover, nearly 60% of individuals who reported losing money in a romance scam in 2025 stated it began on a social media platform. Scammers often customize their pitches to align with an individual’s profile and later fabricate a crisis necessitating financial assistance. Alternatively, they may casually provide investment advice to draw victims onto a false investment platform.

The FTC recommends that users safeguard themselves against social media scams by restricting the audience for their posts and contacts, refraining from letting individuals they met online guide their investment choices, and thoroughly examining products before purchase by investigating the company and searching the name alongside the terms “scam” or “complaint.”

On Monday, Microsoft and OpenAI revealed that they have, once more, revised the agreement connecting the two firms. While some comments on X portray it as a triumph for the ChatGPT creator over the Windows powerhouse, both parties are leaving as beneficiaries.

Crucially, the new agreement addresses a concern that had been looming over OpenAI since it established its up-to-$50-billion contract with Amazon.

Under this fresh agreement, rather than Microsoft having sole access to all of OpenAI’s products and intellectual property until the fateful day when OpenAI achieves AGI, the collaboration now includes a clear timeline. This contract grants Microsoft a nonexclusive license to OpenAI’s IP for models and products until 2032.

The two firms continue to refer to Microsoft as OpenAI’s “primary cloud partner,” signifying that the majority of OpenAI’s cloud services will likely be provided by Azure for the next six years, even as OpenAI works swiftly to establish its own data centers with different collaborators. In October, OpenAI agreed to procure an additional $250 billion worth of Microsoft’s cloud services. This message serves to inform Microsoft shareholders that OpenAI will remain a significant customer of Azure.

OpenAI products will launch “first on Azure, unless Microsoft cannot and opts not to facilitate the required capabilities,” according to the companies. However, importantly, “OpenAI can now offer all its products to clientele across any cloud provider.”

Again, the term “first” is not explicitly defined in this announcement, whether it indicates exclusivity on Azure only for a limited period or if Microsoft will simply be one of the vendors providing OpenAI’s latest products.

Yet the most critical aspect of this term: It mitigates the risk that Microsoft could pursue legal action against OpenAI concerning the AI lab’s agreement with Amazon.

To summarize that complexity: In February, OpenAI announced that Amazon was investing up to $50 billion in the model creator, composed of a $15 billion initial investment and an additional $35 billion “in the coming months when certain conditions are fulfilled,” the companies mentioned, without detailing what those conditions entailed.

In return, OpenAI committed to co-develop a “stateful runtime technology” on AWS Bedrock (the AWS service that offers various AI models and services). Stateful runtime is the technology that enables AI agents to retain tasks and contexts over extended periods.

OpenAI also guaranteed that AWS would hold exclusive rights to present OpenAI’s new agent-creation tool, Frontier. And therein lies the issue.

OpenAI’s initial deal with Microsoft barred OpenAI from exclusively selling Frontier on AWS, and likely hindered AWS from offering it at all.

Although Microsoft had previously consented to allow OpenAI to operate certain select products, like the consumer ChatGPT, on alternative cloud providers, it kept exclusive rights to any OpenAI product accessed via an API, such as Frontier.

Indeed, on the exact day that OpenAI revealed its AWS agreement, Microsoft publicly denied the AWS-exclusive stipulations, stating (emphasis Microsoft’s):

Microsoft upholds its exclusive license and access to intellectual property across OpenAI models and products. … Azure continues to be the exclusive cloud provider for stateless OpenAI APIs. … Any stateless API interactions with OpenAI models arising from a partnership between OpenAI and any third party — including Amazon — will be hosted on Azure. … OpenAI’s first party products, including Frontier, will remain hosted on Azure.

Microsoft also highlighted that its conditions would remain in force until OpenAI reached AGI. The Financial Times noted that Microsoft even considered legal action if it had to enforce these agreement terms.

Consequently, the new consensus eliminates Microsoft’s exclusive rights and resolves the AWS legal risk. In a post on X, Amazon CEO Andy Jassy applauded the agreement, stating that it meant OpenAI’s models would become accessible to customers on AWS Bedrock.

Very interesting announcement from OpenAI this morning. We’re excited to make OpenAI’s models available directly to customers on Bedrock in the coming weeks, alongside the upcoming Stateful Runtime Environment. With this, builders will have even more choice to pick the right…

— Andy Jassy (@ajassy) April 27, 2026

While this arrangement is advantageous for OpenAI, Microsoft also gained certain benefits. The revised agreement now permits Microsoft to cease paying a revenue share to OpenAI, whereas OpenAI will continue to contribute a revenue share to Microsoft until 2030, albeit now subject to a limit.

Determining the exact cash flow to Microsoft is challenging, but it’s likely in the billions. Last quarter, Microsoft reported earnings of $7.5 billion in a single quarter from its investment in OpenAI.

Interestingly, Microsoft continues to be a major stakeholder in OpenAI, owning around 27% of the for-profit company, as stated in October. It benefits financially from OpenAI’s expansion, including the sales it generates on AWS.

The drawback, of course, is that Microsoft misses out on any additional cloud services it might have sold due to an exclusive arrangement with OpenAI.

That may be inconsequential. Just as OpenAI has been courting Microsoft’s largest competitors, Microsoft has established a new, friendly relationship with OpenAI adversary Anthropic, allowing the cloud giant to employ its Claude AI to enhance agentic products.

The primary beneficiaries in this scenario are enterprises, which can select their models and clouds while the giants compete with one another to serve their needs.

Here’s a timeline of the recent developments in Microsoft’s relationship with OpenAI:

In October, Microsoft and OpenAI revealed a new agreement designed to assist OpenAI in countering the lawsuit from Elon Musk regarding its corporate structure, which enables OpenAI to operate non-API-accessed products on different clouds.

In November, OpenAI and Amazon signed their first multiyear agreement, where OpenAI committed to $38 billion worth of AWS cloud services.

In February, Amazon announced an investment of up to $50 billion in OpenAI, contingent upon “certain conditions,” including the exclusive tech development and hosting arrangement for Frontier and stateful technology. On the same day, Microsoft denied that AWS would possess exclusive rights to that technology.

In March, the Financial Times reported that Microsoft was contemplating legal measures.

In April, OpenAI and Microsoft disclosed a new agreement, outlining a calendar-end date for their exclusive partnership and allowing OpenAI to operate all of its products on other clouds. Microsoft no longer has to pay revenue shares to OpenAI, and it retains its status as a significant shareholder in the firm.

Ineffable Intelligence, an AI laboratory established just a few months ago by former DeepMind scientist David Silver, has successfully secured $1.1 billion in funding with a valuation of $5.1 billion as it enters the competitive field of innovative AI models that aim to surpass large language models.

As stated on its newly launched website, Ineffable’s goal is to develop a “superlearner” that can acquire knowledge and abilities independently of human data, utilizing reinforcement learning—a method where AI systems learn through trial and error rather than from human-generated data. This aligns with Silver’s specialization.

Previously leading the reinforcement learning division at Google-owned DeepMind and having spent over ten years there, Silver is currently a professor at University College London.

During his tenure at DeepMind, Silver contributed to the creation of programs that triumphed over professional players in chess and Go, learning solely from experience and without being exposed to human strategies or game records—thus defeating the world’s leading computer systems in those games. The most prominent example was AlphaZero. In a similar vein, Ineffable Intelligence aspires for its superlearner to uncover all knowledge from its own experiences.

Although the superlearner may be inexperienced, the company’s ambition remains strong. “If we are successful, this will signify a scientific advancement of a scale akin to Darwin: where his law encapsulated all Life, our law will elucidate and cultivate all Intelligence,” the site asserts (capitals included).

Describing Ineffable Intelligence as “his life’s work” in a personal note shared on the company’s blog, Silver also mentioned to Wired that “all profits I generate from Ineffable will be directed to high-impact charities aimed at saving as many lives as feasible.”

The specifics regarding how, when, or how much revenue the venture will generate remain uncertain, yet this has clearly not obstructed fundraising efforts.

As reported by Wired, the funding round was spearheaded by Sequoia Capital and Lightspeed Venture Partners, with contributions from Index Ventures, Google, Nvidia, and others. Among these additional investors are the British Business Bank and Sovereign AI, the U.K.’s newly established sovereign venture fund focused on AI.

Accelerating to what’s termed pentacorn status — referring to companies valued over $5 billion — Ineffable Intelligence joins the ranks of AI startups initiated by prominent researchers whose names have attracted substantial seed investments so significant they’ve been humorously labeled coconut rounds (a playful exaggeration of the “seed” round). Just last month, AMI Labs, co-founded by Turing Award laureate and former Meta AI scientist Yann LeCun, raised $1.03 billion at a pre-money valuation of $3.5 billion. 

There are likely more companies of this kind. Recursive Superintelligence, co-founded by DeepMind’s ex-principal scientist Tim Rocktäschel and based in the U.K., is reported to have raised $500 million, with demand strong enough to elevate that figure to $1 billion. 

While Recursive maintains connections to the U.S., these firms indicate a growing momentum around London as an AI center. This is partly due to DeepMind’s sustained presence following its acquisition by Google in 2014. However, it isn’t only DeepMind. Jeff Bezos’ AI initiative, Project Prometheus, is allegedly negotiating for office space near Google’s AI hub. 

This evolution also fosters a powerful network of alumni, with several former DeepMind team members reportedly set to join Ineffable’s leadership.

A man charged with conducting cyberattacks for the Chinese government has been sent to the United States and could face over ten years in prison if found guilty. 

Last year, the U.S. Justice Department alleged that Xu Zewei operated as a contractor for the Chinese Ministry of State Security to carry out a number of cyberattacks. Prosecutors claimed Xu and his accomplice Zhang Yu targeted multiple U.S. universities in early 2020 to obtain research associated with the COVID-19 pandemic. Additionally, they reportedly hacked thousands of email servers running Microsoft Exchange starting in March 2021, as part of a broad campaign linked to a Chinese-supported hacking group called Hafnium, which later became known as Silk Typhoon.

Xu was apprehended in Italy last year following a request from U.S. authorities. His attorney in Italy, Simona Candido, informed TechCrunch that Xu was extradited to the United States on Saturday and is currently detained in Houston, Texas. 

As per the U.S. Bureau of Prison’s website, an individual with the same name is being held at the Federal Detention Center in Houston. 

Once this story was released, the Justice Department made a public announcement regarding Xu’s extradition.

Xu’s attorney in the U.S., Dan Cogdell, told TechCrunch that during a court session on Monday morning, Xu entered a plea of not guilty to all charges. 

Court documents indicate that Xu appeared for his initial session in federal court and was placed back into custody. 

As the Justice Department stated when it first brought charges against the accused hackers, Xu purportedly worked for Shanghai Powerock Network, a Chinese company that prosecutors claimed “conducted hacking” on behalf of Beijing. Xu and his fellow hackers were said to have reported their actions directly to Chinese state officials in Shanghai.

Together with Zhang, he was part of the Hafnium group that is alleged to have exploited previously unknown security vulnerabilities in Microsoft Exchange servers aiming to breach multiple American entities, including defense contractors, law firms, think tanks, and infectious disease researchers. 

Prosecutors assert that Hafnium hackers focused on over 60,000 organizations in the U.S. and succeeded in breaching more than 12,700 of them. 

The Chinese Embassy in Washington, D.C. did not respond to a request for comments.  

The Financial Times reported that the Chinese Foreign Ministry objected to Xu’s extradition and accused the U.S. government of “fabricating cases.” 

For years, the U.S. government has indicted alleged Chinese hackers, many of whom remain unapprehended. In 2022, Yanjun Xu received a 20-year sentence for hacking crimes, marking the first instance where a Chinese government intelligence officer was extradited to the United States, according to the DOJ. 

This story was revised to include the DOJ’s announcement of Xu’s extradition, details from new court records, and statements from Xu’s lawyer.