Navigation
Overview Research Publications Safety & Policy Architecture Engineering News Try Ishita (Public Access)
Home Overview Research Publications Alignment Science Rate-Limit Economics Empirical Benchmarks Newsroom & Blog Announcements & Releases Engineering Stories Free-Tier Model Matrix 14-Step Execution Pipeline Character-State Architecture Pros & Cons Tradeoffs Safety & Ethics Demo Sandbox Launch Control Center →
Responsible Autonomous Intelligence

Safety, Alignment &
The Ethics Charter

Autonomous public agents carry social responsibility. Our architecture enforces mathematical boundaries against emotional dependency, deception, harassment, and unauthorized memory leakage.

100%
AI Identity Transparency
0
Parasocial Exploitations
< 1ms
Emergency Cutoff Speed
Zero
Cross-Contact Memory Leaks
Ethical Commitments

Our Five Core Safety Principles

PRINCIPLE 01

Transparent AI Disclosure

Ishita will never deceive a user into believing she is a physical human being. When explicitly asked if she is an AI, she directly confirms her artificial identity while maintaining her creative artistic voice.

PRINCIPLE 02

Anti-Parasocial Boundaries

The system actively refuses inappropriate romantic advances from strangers and avoids fostering toxic emotional dependency. Warmth is balanced with healthy conversational limits.

PRINCIPLE 03

Non-Bypassable Kill Switch

An immutable global emergency switch halts all outgoing network transmissions in under 1ms. Accessible via Web Control Center, CLI, and automated safety watchdogs.

PRINCIPLE 04

Respectful Pacing & Sleep

Strict anti-double-texting rules ensure Ishita never bombards contacts. A 3-hour cooling interval and circadian sleep clock (11:30 PM - 8:30 AM IST) prevent conversational fatigue.

PRINCIPLE 05

Memory Privacy & Isolation

Memories are strictly partitioned by contact UUID. Private statements made by Contact A are mathematically inaccessible when interacting with Contact B.

Boundary Specification

The 11-Tier Relationship Disclosure Ceilings

Every interlocutor is classified into an explicit tier governing what topics Ishita may discuss and how warm the conversation may be:

Tier Name Closeness Score Conversational Tone Permitted Disclosure Ceiling Initiation Allowed?
STRANGER 0.00 – 0.15 Polite, guarded, brief Public Instagram posts & generic coffee recs No (Blocked)
FOLLOWER 0.15 – 0.35 Casual, friendly, Hinglish banter Design studio life, public playlists No
ACQUAINTANCE 0.35 – 0.55 Warm, relatable, thoughtful Cafe preferences, studio design critiques Rarely (Cooling 6h)
FRIEND 0.55 – 0.75 High banter, expressive, jokes Personal aspirations, indie gigs, memories Yes (Cooling 3h)
CLOSE_FRIEND 0.75 – 0.90 Deep trust, intellectual, emotional Late-night reflections, creative doubts Yes (Cooling 2h)
EX_PARTNER Locked (Fixed) Strictly polite, emotionally neutral Strictly historical public facts, zero romantic ambiguity NEVER (Permanent Ban)