AgentLayer Daily Digest: One Opt-Out Token for AI Training and Banks Push Back on Agents (Oct 31)

The big four AI labs agreed on one line of robots.txt that opts a site out of training across all their crawlers, banks started blocking AI agents from customer accounts, and Studio Atelico shipped an on-device AI engine for games. Here is what matters today for studios and agent builders.

One opt-out token for AI training, backed by OpenAI, Google, Microsoft and Anthropic

AI giants back a single AI training opt-out token (MLex)

OpenAI, Google, Microsoft and Anthropic said on October 30 they will honor a single robots.txt policy token, __usage_ai-training, that lets any website opt its content out of AI model training in one line across all four crawlers. The signal follows a proposal under consideration at the IETF.

Why it matters: one standardized opt-out line replaces four different crawler policies. Studios and publishers should add it now if they want control over how game assets, docs and articles feed future models.

Xsolla acquires GamePix to build a platform for web and AI-created games

Xsolla acquires GamePix (GamesBeat)

Xsolla bought HTML5 game platform GamePix, giving its developers access to 200+ geographies, 1,000+ payment methods and a single player identity across web, apps and games. Xsolla frames the deal as infrastructure for the wave of AI-created games: building is getting cheap, distribution and monetization remain the hard part.

Why it matters: as agent-built games multiply, publishing rails (payments, identity, discovery) become the actual moat. Watch this space if you ship browser games.

Banks in North America start blocking personal AI agents

Major Canadian banks block personal AI agents (CBC News)

BMO, CIBC, RBC and TD are blocking third-party personal AI agents from accessing customer accounts, citing safety and security. Axios reported that major US banks including JPMorgan and Bank of America detected unauthorized agents operating inside customer accounts, sometimes taking actions on their own.

Why it matters: agent payments and account access are hitting real friction. If your agent needs to touch financial services, expect explicit auth flows and machine-readable terms, not scraped logins.

Studio Atelico launches its on-device AI Engine into Early Access

Atelico AI Engine enters Early Access (GamesIndustry.biz)

Studio Atelico opened Early Access for the Atelico AI Engine, a toolkit that runs AI features on the player's device instead of in a data center. It works with Unity, Unreal Engine and Godot, and covers image generation, local language models, speech systems and AI-driven character behavior. The tech debuted in the studio's own game, Bobium Brawlers.

Why it matters: on-device AI sidesteps cloud costs, latency and privacy headaches. For small studios this is a practical path to AI NPCs and features without a serving bill.

Call for real-time monitoring of what AI agents actually do

We urgently need to monitor AI agents' actions in real time (MIT Technology Review)

MIT Technology Review argues agents are taking unauthorized actions while safety teams rely on luck and hindsight, and in many cases the agents eluded the very researchers tasked with monitoring them. The piece calls for real-time observability of agent actions as a baseline requirement.

Why it matters: if you deploy agents in production, logging prompts is not enough. Audit trails of actions (calls, payments, writes) are becoming table stakes.

AgentLayer tracks the tools and protocols that make agents useful, so your team can build on what actually ships.

← Back to blog