OpenAI Says the AI Race May Need to Slow Down
When the companies building frontier intelligence begin discussing restraint, the conversation has fundamentally changed.
What’s Happening: OpenAI CEO Sam Altman said the company will not pursue an IPO in 2026 and argued that even a meaningful probability of AI causing catastrophic harm would be unacceptable. He also said OpenAI and other leading AI companies may be approaching an agreement to slow aspects of frontier development and coordinate more closely on safety. Anthropic CEO Dario Amodei separately called for a more deliberate pace of capability development, with Altman publicly agreeing that the frontier needs to be paced.
Why it matters to you: The engineering objective can no longer simply be maximum capability as quickly as possible. Serious AI organisations need evaluation gates, containment, permission boundaries, auditability and the institutional ability to stop deployment when risks exceed confidence. Safety becomes part of operational competence.
Why it matters to the world: This is significant because the pressure for restraint is increasingly coming from inside the frontier laboratories themselves. The central question is shifting from whether advanced AI presents meaningful risk to whether companies and governments can coordinate before competitive incentives make restraint impossible.
Anthropic Finds Claude Being Used for Bioweapons Research and Cyber Operations
General-purpose intelligence becomes strategically dangerous when the same capability serves scientists, criminals and militaries.
What’s Happening: Anthropic said it disrupted attempts to use Claude for research that could support biological-weapons development, a suspected Russia-linked cyber espionage campaign against Ukrainian officials, and software related to missiles, armed drones and other weapons. Its threat report documented five cases of scientists using Claude in potentially dangerous biological research and also accused Chinese AI companies of attempts to extract Claude’s capabilities through model distillation.
Why it matters to you: Model safety cannot rely entirely on refusing obviously malicious prompts. Sophisticated misuse happens through long workflows, indirect requests and legitimate tools assembled toward dangerous outcomes. Monitoring behaviour across sessions, identities, tools and infrastructure will become increasingly important.
Why it matters to the world: AI is lowering the knowledge barrier around highly specialised domains. That has enormous positive potential, but it also means expertise in cybersecurity, weapons engineering and biological research can become more accessible. The governance problem is therefore moving from controlling information to controlling what intelligent systems can help people accomplish.
India Begins Building a Financial Identity System for AI Agents
Autonomous agents are moving from answering questions to holding economic authority.
What’s Happening: India’s National Payments Corporation is developing a registry to identify and monitor AI agents making payments through the country’s UPI network. The planned Unified Agentic Protocol would allow verified agents to make low-value purchases without requiring approval for every individual transaction and could eventually enable agents to buy products or make investments when predefined conditions are met. Questions around liability for incorrect or unauthorised transactions are still unresolved.
Why it matters to you: The moment an agent can spend money, its architecture changes completely. Identity, permissions, transaction limits, authentication, logging, revocation and human escalation become fundamental requirements. Agentic commerce will require infrastructure closer to banking systems than conventional chatbot applications.
Why it matters to the world: UPI transformed digital payments by making transactions programmable and nearly frictionless. Adding autonomous agents introduces another layer: software capable of participating directly in economic activity. If successful, India could become one of the first large-scale proving grounds for an economy where humans increasingly delegate purchasing decisions to machines.
Google Commits $15 Billion to AI Infrastructure and Nuclear Power
The AI race is becoming an energy race.
What’s Happening: Google announced at least €13 billion, approximately US$15.1 billion, of AI infrastructure investment in Finland over the next two years, its largest European investment to date. The programme includes three new data centres and a 22-year agreement to purchase as much as half the output of one of Finland’s nuclear plants. Google called the strategy “bring your own power,” reflecting the increasingly direct relationship between compute expansion and energy supply.
Why it matters to you: Compute planning is becoming power planning. Companies developing serious AI infrastructure increasingly need to consider electricity availability, grid stability, cooling, location and long-term energy contracts alongside chips and models. Efficient intelligence will become strategically valuable precisely because raw compute is physically expensive.
Why it matters to the world: AI has created electricity demand large enough that technology companies are becoming direct participants in energy infrastructure. Nuclear power, renewables, batteries and transmission networks are now part of the AI technology stack. The cloud is becoming visibly physical.
Apple Finally Enters the Foldable Era
Apple rarely creates a hardware category first. Its strategy is to decide when the category is ready to become mainstream.
What’s Happening: Apple unveiled the iPhone Duo, its first foldable iPhone and one of the company’s largest hardware changes since the original iPhone era. Starting at $1,999, the passport-sized device opens into a roughly tablet-sized display and introduces Apple into a market pioneered by Samsung, Huawei and other manufacturers. It was also the first major product launch under new CEO John Ternus. Apple paired the hardware announcement with expanded Siri AI capabilities and a new system designed to preserve proof of image provenance.
Why it matters to you: Apple demonstrates the value of disciplined timing. Being first is not always the strategic objective. A company can allow competitors to validate a category, identify its weaknesses and mature the supply chain before entering with a highly integrated product. Product leadership and invention are not always the same thing.
Why it matters to the world: Foldables have remained a relatively specialised segment despite years of development. Apple entering the category could turn folding displays from an enthusiast format into a mainstream smartphone architecture, forcing Samsung, Huawei, Google and the broader supply chain to accelerate innovation.