Gabriel Odeyemi
← Strategic Digest

Strategic Digest

Who Sets the Defaults Wins the Decade

As Europe pries open Google and China courts the developing world, the fight over agentic AI has become a contest over control points, not capability.

Gabriel Odeyemi · · 6 min read

On Thursday the European Union told Google it must open key parts of Android and Google Search to rival AI assistants and search engines. The decision, handed down under the bloc's digital antitrust rules, is the first regulatory action aimed squarely at the layer where AI assistants reach users. It arrived within hours of Xi Jinping describing artificial intelligence as a "symphony of global collaboration." The juxtaposition is not coincidence so much as coincidence made meaningful. The West is tightening the rules on its own champions at the exact moment China is positioning itself as the permissionless alternative. The contest that matters now is not who builds the most capable model. It is who controls the defaults.

The Distribution Layer Becomes the Battleground

The EU order matters less for what it does to Google's balance sheet than for where it strikes. By forcing greater access to Android and Search, regulators are opening the distribution layer through which AI assistants meet users at the point of default. That is the scarce real estate of the coming decade. A model can be excellent and still invisible if it cannot reach a user without paying an incumbent's toll.

The ruling could weaken Google's control over two of the industry's most important platforms. For anyone whose growth strategy assumes reaching European users through Android or Search, the acquisition math has shifted. The competitive opening exists now, while rivals are still mobilizing. It will not stay open indefinitely, and incumbents rarely cede a channel without building a new one.

What makes this more than a routine antitrust action is timing. Google is simultaneously pushing its own assistants deeper into everything it owns, letting AI Mode link and act across select apps, renaming NotebookLM to Gemini Notebook, and adding personalized avatars to Vids. The company is integrating; the regulator is prying open. Both are betting on the same premise: the assistant that owns the default owns the relationship.

Software Rebuilt for Machines, Not People

The second shift is quieter and arguably larger. Software is being rebuilt for AI actors rather than human ones. Anthropic's Claude can now use stored 1Password credentials to complete multi-step tasks such as booking travel and managing online accounts on a user's behalf. DoorDash has opened a limited beta of a command-line tool that lets developers and AI agents search stores, build carts, and place orders from a terminal.

Neither headline is dramatic on its own. Together they describe an interface being reconstructed for machine users. When an agent holds live credentials and can transact, the assumptions that governed the consumer web start to fail. Who is liable when an agent authorizes a purchase no human reviewed? Who bears the loss when stored logins are misused? These are not abstract questions. They are the terms on which the agentic web will run, and they are being set now, largely by whoever ships first.

The prudent response is unglamorous. Firms should decide, before an employee does it for them, whether AI agents may hold company logins. The first major security breach or disputed transaction involving an autonomous agent will force a liability reckoning, and organizations that have not written a policy will be writing one under pressure.

China's Open Gambit and the Closing Space Gap

Xi's "symphony" language is a strategic wedge, not a philosophy. As the West regulates its incumbents and litigates its own rules, Beijing is courting the developing-world AI market as the champion of open standards and frictionless adoption. It is a shrewd contrast: the United States and Europe optimizing for safety and antitrust, China optimizing for reach.

The positioning gains force from a hard capability signal. On July 10, China Aerospace Science and Technology Corp. launched the Long March-10B on its first flight, placed its payload into orbit, and then caught the returning booster midflight, a feat only SpaceX had managed. China uses nets where SpaceX uses "chopsticks," but the economics are the point. Reusable launch is the cost advantage that underwrote American commercial-space dominance, and Beijing is compressing it. Analysts quoted in the coverage suggested this should worry Elon Musk more than any single rival has before.

When a country can credibly claim both the moral high ground on openness and the engineering parity on frontier hardware, its pitch to undecided markets grows more persuasive. The West's internal rulemaking, however necessary, reads from the outside as friction.

Capability Racing Ahead of Its Guardrails

Running underneath all of this is a widening gap between what AI can do and the controls meant to contain it. OpenAI has built an internal LLM super-hacker it calls GPT-Red, using it as a sparring partner to make its other models safer. The tool is a candid admission that capability now outruns assurance, and that the only credible test of a system is another system.

The exposure extends beyond models to the data they consume. The risk of weather-data sabotage is rising, and forecasts quietly underpin decisions across grids, aviation, and agriculture. A poisoned source could cascade into operational failures before anyone detects the manipulation. As agents act on data at machine speed, the integrity of the underlying inputs becomes a first-order security concern rather than a technical footnote.

The same week offered a smaller signal worth noting. A new bottled-water brand, Loonen, is selling verified purity, water filtered of microplastics and forever chemicals, as a premium good. Consumer safety is becoming a status symbol, a template likely to extend from water to food, air, and eventually data provenance. When trust becomes scarce, verified trust becomes a market.

The Strategic Read

The defining tension of the moment is that AI is moving from answering questions to taking consequential actions, and no one has settled who is liable or who controls the choke points. Europe is fighting over rules, the United States is building red-team defenses, and China is selling openness while closing the hardware gap. The pattern favors whoever sets the defaults for agentic AI: the credential access, the app integrations, the distribution channels, the launch economics.

Three moves follow. Reassess any strategy that assumes Google-default distribution in Europe, because the EU ruling has changed the math. Convene a short, serious review of agentic-AI exposure before an employee authorizes an agent to hold company logins without approval. And stress-test near-term energy assumptions against sustained Gulf disruption, where shipping at a one-month low and rising oil prices are an early indicator rather than a peak. The West is optimizing for safety and antitrust. China is optimizing for adoption and capability. The winners of the next decade will be decided at the control points, not the demos.

Sources

← Back to all analysis