Dear Reader,

As July draws to a close, the fragile détente on the global stage has begun to unravel. Full-scale war has returned to West Asia, despite a growing consensus that further military escalation is unlikely to resolve the conflict and will only inflict greater suffering on ordinary people. Meanwhile, the US administration has embarked on a renewed wave of economic aggression, with Donald Trump resorting to a legal workaround to reinstate his tariffs after the Supreme Court had previously ruled them unconstitutional.

As before, these tariffs appear to be wielded as a blunt instrument to restore the US’s fraying authority, with the tech sector singled out as a key target. Trump himself made this explicit this month, openly threatening to raise tariff rates on the EU in retaliation for the fines its regulators have imposed on major technology companies. Yet, given that similar efforts were undone by judicial scrutiny in the past, and that the legal justifications advanced this time appear equally — if not more — tenuous, there is good reason to doubt that this strategy will prove any more effective today.

However, one can understand why such a need to lash out is being felt in the US, as recent developments illustrate the growing headwinds confronting Silicon Valley’s dominance. This month, for instance, witnessed what many have described as another ‘DeepSeek moment.’ The release of Beijing-based Moonshot AI’s new generative AI model, Kimi K3, has sent ripples through the industry. Not only has the model been assessed as competitive with the leading frontier models developed in the West, but it also appears to have achieved this at a significantly lower cost than its rivals. Moreover, by being both free and open source, it has raised serious questions about the expensive, proprietary business model favoured by US firms.

Predictably, the release also triggered a sharp correction in US technology stocks, with nearly $400 billion wiped from the combined valuations of OpenAI and Anthropic. The episode has also reignited geopolitical concerns, with some analysts arguing that the US sanctions regime has failed to prevent advanced chips from reaching Chinese AI start-ups, and that a far more aggressive strategy is now required.

Moreover, US officials have accused the Beijing-based company of intellectual property theft, arguing that its open-source model derived its capabilities by using the outputs of frontier models such as Claude and Fable Five. However, as critics have noted, questions of intellectual property are particularly murky in this domain, given the way AI models themselves have been trained. Just this month, Anthropic reached a landmark $1.5 billion settlement in a class-action lawsuit brought by authors who alleged that the company had infringed their copyrights by using their works as training data. Reportedly the largest settlement of its kind in a US copyright case, it is likely to pave the way for further litigation. At the same time, the settlement weakens Anthropic’s ability to accuse open-source models trained on Claude of infringement, given the contested intellectual property foundations of Claude itself. How this increasingly tangled legal landscape evolves will be an important issue to watch in the months ahead.

Notably, this was not the only troubling AI-related development in recent weeks. An investigation by Nikkei Asia found that five of the largest US technology companies — Meta, Alphabet, Amazon, Microsoft, and Oracle — have concealed an estimated $1.65 trillion in debt from their balance sheets, much of it linked to AI-related investments. This is a remarkable sum. According to the investigation, these companies have relied on subsidiary entities and other off-balance-sheet financial arrangements to obscure the scale of these liabilities. The findings have fuelled concerns about the fragility of the AI bubble, prompting comparisons with Enron, whose collapse was precipitated by similar hidden debts.

These concerns have been further amplified by new reporting on the central role that private credit is playing in financing AI investment. Rather than relying on traditional banks, many firms are borrowing from non-bank lenders such as insurance funds and asset managers. This creates a far more opaque financial landscape, in which the terms of lending are less transparent and underwriting standards may be considerably weaker. At a moment when the global economy is already grappling with supply shocks and anti-AI sentiment is becoming increasingly widespread, Silicon Valley — and perhaps the American economy more broadly — appears to be increasingly dependent on the one industry that is critically vulnerable to a major financial correction. That only adds to the volatility of the present moment.

Amid this turbulence, countries across the Global South are accelerating efforts to reduce their technological dependencies and build digital infrastructures that can secure greater autonomy in the face of mounting US pressure. This month, a series of developments highlighted the central role that digital payment systems play in this process. India announced the expansion of its Unified Payments Interface (UPI) into Europe, launching operations in Greece and France. The system is now live in around a dozen countries, with further international expansion planned over the coming years.

Meanwhile, Brazil’s central bank has, over the past few months, signed information-sharing agreements with more than 65 countries seeking to develop payment systems modelled on its own Pix platform. Pix’s growing international influence has itself become a point of friction with Washington, with some analysts identifying it as one of the factors behind the harsher treatment Brazil has received in Trump’s latest round of tariffs.

Parallelly, July also saw Tanzania and Rwanda advance an initiative to develop an integrated cross-border digital payments system. Drawing on successful models such as Pix and UPI, the initiative aims to create a regional payments network that reduces reliance on the infrastructure of international operators such as Visa and Mastercard.

Taken together, these developments point to the emerging common sense across the Global South: the digital age demands the cultivation of local technological capabilities and sovereign digital infrastructure, as dependence on foreign platforms and payment rails increasingly carries the risk of economic subordination.

Finally, this month also marked another important front in the Global Majority’s struggle for self-determination with the publication of the main framework draft for the UN Tax Convention, ahead of the next round of negotiations scheduled for August. This process could prove to be the most significant development in the history of international tax cooperation, with the longstanding tax avoidance strategies of multinational technology companies forming one of the central issues the convention seeks to address.

According to reports from the last round of negotiations, resistance continues to come primarily from OECD countries. This is particularly striking given that many of them were themselves among the principal casualties of the hollowing out of the OECD’s earlier Global Tax Agreement, after sustained corporate lobbying and political manoeuvring diluted the agreement and secured major exemptions for US multinationals. As with so much in global politics today, the outcome may ultimately hinge on a broader test of US hegemony: whether Washington’s allies are prepared to break with an increasingly fragile order. If they find the courage, the possibilities could be extraordinary.

The Sins & Synergies Lounge 

Check out this interactive art project “to hurt for a thought” that materializes the ecological costs of AI in a heating metal keyboard.

Read this analysis of the challenges in governing India’s subsea cable ecosystem and an argument for treating it as critical national infrastructure.

Also read Jovan Kurbalija’s take on how the UN may not be the fastest actor in the AI race but that this slow, deliberate, inclusive, and patient approach might just lead to better governance.

How can the Wikimedia Movement retain its agency in the new AI knowledge loop? Check out this paper by Open Future on the ongoing structural transformation of the knowledge ecosystem to help articulate a direction for the Wikimedia Movement’s response.

Don’t miss this thoughtful deep dive into what it would take for municipal governments to build their own AI models.

Finally, explore Dirty Data, an initiative keeping a close watch on the development of data centers worldwide and its human, ecological, and infrastructural costs.