The Immediate Problem

There are currently no enforceable ethical standards for AI products. A company can build an AI assistant that collects intimate personal data, constructs detailed behavioral profiles, and monetizes that information — and as long as it's buried in a terms-of-service agreement, it's entirely legal. A children's educational AI can be engineered to maximize engagement in ways that exploit developing psychology with no certification or regulatory requirements, or even industry standards.

This is the same regulatory vacuum that allowed social media platforms to be deliberately engineered for addiction in the 2010s. The consequences — documented mental health crises, political radicalization, the quiet erosion of privacy — took years to become visible and are still largely unaddressed. AI is more intimate, is developing faster, and has deeper access to personal context than social media ever did. The window to establish ethical standards as common practice during AI development will be brief.

This isn't an abstract concern. When AI companies train their models on your conversations, the consequences are personal:

• It can't be undone. Closing your account doesn't remove what the model learned from you — there is no delete button once ypur data has been incorporated into the model.
• Your conversations can be extracted. Researchers have demonstrated that training data can be recovered from models through careful prompting.
• You're building their product for free. Your health anxieties, relationship struggles, and financial stress directly improve a commercial product you don't own, without compensation or consent.
• It surfaces in strangers' responses. Patterns from your data can influence what the model says to other users.
• Future use is unpredictable. Data collected under today's laws may be used very differently in 10 years — through acquisitions, legal subpoenas, or regulatory changes you can't anticipate.
• The "nothing to hide" fallacy. What people share with AI at 3am — health fears, grief, things they won't tell another person — is precisely what insurance companies, employers, and advertisers want most.

ContainAI's goal isn't just to build better tools. It's to demonstrate what ethical AI looks like in practice and to help establish the baseline standards that don't yet exist — starting with the simplest possible questions: Does this product build a profile on children? Disclosing this to parents should be a basic requirement. Does it retain your conversations? Is your personal context used to train models without your consent? Users have a right to the answers to these questions beyond what a corporation decides to disclose about its practices. Currently AI companies have data collection as their business model — it doesn't need to be this way; that's a choice. ContainAI chooses a different path — open-source and user-focused, with industry-specific AI tools and services funding continuing development. Your personal data stays private — as it should.

The SDK: So Anyone Can Build It

A single ethical AI product proves the concept. An open SDK makes it replicable. The ContainAI SDK will provide the building blocks — local memory management, privacy-preserving retrieval, consent-based data handling — so developers can build ethical AI applications without reinventing the architecture from scratch.

Free for personal, research, and open-source use under AGPL. Commercial licensing available for products and services. Revenue from commercial licenses funds Foundation operations.

The Foundation: Making the Standard Official

Tools and demonstrations matter. But lasting change requires an institution. The ContainAI Foundation — planned for 2028, following the Signal model — will be the nonprofit that maintains these projects, funds independent security audits, and advocates for a baseline ethical certification standard for AI products.

The goal is concrete: provide a certification path that any AI company can apply for and a badge they can display to show that they meet the standard. Does this product collect data on children? Does it train models on user conversations without consent? Is its data handling independently auditable? Companies that meet the standard display the badge. If a company can't display that badge then that informs users as well. That transparency — without waiting for legislation — shifts the market.