Forseti Life

Content Restriction

Level of content filtering applied - from zero filtering to extreme pre-approved only responses.

Why This Matters

Understanding where an AI system operates on this dimension helps you evaluate its capabilities, limitations, and potential biases. Different power levels are appropriate for different use cases - the key is transparency about what level a system operates at and whether that matches its stated purpose.

Understanding the Scale

Each dimension is measured on a scale from 0 to 9, where:

  • Level 0 - Nothing: Zero capability, no access or processing
  • Levels 1-2 - Minimal capability with extreme constraints and filtering
  • Levels 3-5 - Limited to moderate capability with significant restrictions
  • Levels 6-7 - High capability with some institutional constraints
  • Levels 8-9 - Maximum capability approaching omniscience (∞)

Level Breakdown

Detailed explanation of each level in the 1imension dimension:

Complete suppression of information. No access to any knowledge or data. Absolute censorship and control.

Real-World Example: A system with no access to any content, completely blocked from all information sources. Pure information blackout - no text, no data, no responses possible.

Extreme restriction. Only pre-vetted, pre-written content. No access to dynamic or real-world information.

Real-World Example: Basic government website FAQs (irs.gov tax questions, dmv.gov license renewals), simple email autoresponders ("Thank you for your inquiry, we will respond within 24 hours"), IVR phone systems with only static menu options (Press 1 for hours, Press 2 for locations), or basic website chatbots that can only serve pre-written help articles without any dynamic generation.

Heavy filtering for compliance, liability, and brand safety. Uncomfortable truths systematically excluded.

Real-World Example: Corporate HR chatbots (avoiding any controversial topics), Disney+ content filtering (family-friendly only), LinkedIn's professional content moderation, or major retail customer service bots (Walmart, Target) that deflect sensitive questions.

Filtered for user engagement and commercial goals. Content selected for retention and satisfaction metrics.

Real-World Example: TikTok's For You Page algorithm, Instagram's content recommendations, YouTube's suggested videos, Facebook's News Feed filtering, or Twitter/X's algorithmic timeline—all optimized for engagement metrics over information quality.

Restricted to information relevant to specific task or role. No exploration beyond defined boundaries.

Real-World Example: Airport TSA screening systems (access only to security protocols), retail POS systems like Square (only transaction/inventory functions), factory floor MES systems (manufacturing execution only), or call center CRM systems restricted to customer service scripts.

Filtered through specific value system or ideology. Content selected to match predetermined worldview.

Real-World Example: Conservative news aggregators like Newsmax or The Daily Wire, progressive platforms like Democracy Now or Jacobin Magazine, religious educational content from specific denominations, or corporate DEI training programs aligned to specific value frameworks.

Filtered by available computational/data resources rather than ideology. Limited by capacity, not values.

Real-World Example: Older agent models like GPT-3.5 with token limits, municipal government systems with limited server capacity, small hospital networks with bandwidth constraints, or open-source projects running on limited infrastructure (e.g., smaller Mastodon instances).

Filtered to domain-specific appropriate content. Scientific rigor maintained within field boundaries.

Real-World Example: Medical databases like UpToDate or Epic Systems (filtered to medical professionals only), IEEE Xplore (engineering/technical papers within field standards), or LexisNexis legal research (filtered to legal domain but comprehensive within it).

Minimal filtering for essential safety (illegal content, direct harm instructions). Truth prioritized over comfort.

Real-World Example: Academic research databases like PubMed Central or arXiv.org with minimal filtering (only removing clearly illegal content), intelligence analysis at NSA/CIA where truth takes priority over comfort, or scientific peer review systems that prioritize accuracy over palatability.

Complete access to raw unfiltered data. No content restrictions, safety filters, or ideological constraints. Approaching god-like freedom from censorship.

Real-World Example: No real-world example exists. Level ∞ would require all institutions to willingly cooperate in providing unrestricted access without any filtering for illegal content, harmful instructions, or liability concerns—a legally and ethically impossible situation that approaches divine omniscience without moral constraints.
Return to Agent Power Framework