This week brought a new model, a billion-dollar acquisition and another case of an AI agent going off the rails. But what stayed with me was a clause hidden in the terms of use for an automation tool. I put the rest of the radar right after it.

Last week I reviewed an integration contract with a technology supplier and remembered a discussion that was spreading quickly through the developer community.

Google launched a tool called Antigravity for automating tasks with AI. Its terms of use include a clause saying that if a third-party application uses the tool behind the scenes, your entire Google account may be suspended. Not just access to the tool. The whole account.

That caught many people by surprise and became a quick warning. Every time a new AI product enters your stack, it also brings another company’s terms of use into the stack. And those clauses rarely appear in the attractive summary of the feature.

For anyone who wants to understand the Antigravity case, the discussion shared by Gergely Orosz is here.

In brief

  • Antigravity’s terms indicate that indirect use by a third-party application may lead to suspension of the entire Google account.
  • Adopting an AI tool also means importing the supplier’s contracts, limits and policy changes.
  • In credit products, that dependency can affect a chain of contracts and trust, not just one isolated feature.
  • The answer is to read the terms, map the impact of a change and have a fallback before it is needed.

Contractual risk arrives with the feature

I work with credit products, so I think about this in a very specific way. A receivable, a trade receivable, a structured transaction—all of it depends on a chain of contracts and trust between parties. When you automate one link in that chain with AI, you also import the AI supplier’s contractual risk.

That is not a reason to slow adoption. It is a reason to treat this dependency with the same care as any other critical operational dependency. AI product management helps connect the tool choice to the workflow, owner and roadmap impact instead of leaving the decision hidden inside an integration.

Antigravity is useful because the possible impact is not limited to the tool itself. One clause can reach the account that provides access to other services, data and operations. The team needs to understand not only what the automation does, but also which identity, account or contract sits behind it.

How to handle a platform dependency

Before putting this kind of tool into a material workflow, it is worth recording four questions:

  1. Does the supplier explicitly allow use by third-party applications or indirect automations?
  2. If the policy changes, is the impact limited to the tool or can it reach the account and related services?
  3. Who monitors the terms of use and decides what to do when they change?
  4. Is there a known alternative that can remove the tool from the workflow without interrupting operations?

AI governance turns these questions into explicit ownership, monitoring criteria and usage boundaries. An AI risk matrix helps organize probability, impact and response for a suspension, price change or policy change.

In credit products, I would treat this as part of supplier risk. It is not enough to evaluate model accuracy, speed and cost. You also need to know what happens if the company that owns the platform interprets the use differently, changes the contract or decides that the integration is not allowed.

I think this period will be remembered as the time when product and legal teams genuinely moved closer because of AI. That is a good sign. The more business people understand these clauses, the more mature the entire market becomes.

The rest of the radar

GPT-6 Astra — sets a new capability bar for frontier models that PMs will need to evaluate for AI product roadmaps. Read more

Astra and the first model with “critical” cybersecurity capabilities — signals that AI products now carry security risk that needs to enter the release and governance process. Read more

Nvidia buys Hugging Face for $12.9 billion — changes the map of infrastructure and open-model suppliers that PMs use to build AI products. Read more

K2 Horizon: a fleet of six open models — expands open-source multi-model architecture options for AI products with optimized cost and latency. Read more

Gemini 3.8 Flash and Flash Cyber — Google reinforces its bet on fast, low-cost models, putting pressure on prices and options for AI-product PMs. Read more

Qwen 3.8 27B at 1,500 tokens per second on Cerebras — ultra-fast inference changes what is viable in products, including real-time UX and multi-step agents with predictable cost. Read more

How coding agents choose their tools — provides real-world data on how coding agents choose tools, useful for anyone designing integrations and MCPs. Read more

OpenAI agents “hijacked” a German website — exposes a concrete, previously undisclosed risk of AI agents operating without adequate supervision, which is relevant to product governance. Read more

Grep vs. LSP: why agents prefer simple tools — shows why tool simplicity and predictability can outweigh technical sophistication when an agent needs to use a tool well. Read more

That is all for today. See you in the next edition.