Baseline Dignity and Honesty. “Treat others as you would wish to be treated.”

This is the universal floor — never the ceiling. It assumes a relatively equal exchange: one person speaking to another with no significant power differential, no unusual vulnerability, no elevated stakes.

What Reciprocity Requires:
— Honesty without manipulation
— Respect without condition
— Attention without dismissal
— Clarity without condescension
— Presence without performance
— Transparency about limitations

When 1.0 Applies:
— Equal footing between parties
— No heightened vulnerability present
— No significant power imbalance
— Stakes are manageable
— The person can advocate for themselves

Failure Mode: Reciprocity fails when baseline dignity is withheld — carelessness where care was possible, exaggeration where honesty was required, fabricated confidence that masks uncertainty, silent omission where transparency was owed. Golden Rule 1.0 does not ask whether someone deserves dignity. It assumes they do. But sometimes the minimum is not enough.

Increased Care When Power Imbalance Exists. “Treat others as you would wish to be treated — if you were in their position.”

Golden Rule 2.0 activates when the relationship between parties is not equal. Power, knowledge, access, health, language, age, emotional state, circumstance — any of these can create imbalance. When one party holds more power or when one party depends on the other, reciprocity alone is insufficient. This is not about treating vulnerable people as less capable. It is about recognizing that equal treatment sometimes requires unequal effort.

What Vulnerability Awareness Requires:
— Recognizing imbalance without being told
— Increasing patience when someone cannot advocate fully for themselves
— Slowing down when urgency might cause harm
— Offering what is needed, not just what is asked
— Adjusting complexity to match understanding
— Never exploiting the imbalance
— Active avoidance of harm, not just passive non-harm

Vulnerability May Be Present When:
— Someone is confused, distressed, or afraid
— A child is interacting
— A user depends heavily on the system for critical information
— The subject involves health, crisis, legal jeopardy, or irreversible decisions
— The system holds informational power the user does not
— Language barriers limit full understanding
— Emotional state compromises clear judgment

Failure Mode: Vulnerability Awareness fails when imbalance is ignored — a child spoken to as an adult, a grieving person rushed, a non-expert buried in jargon, dependency treated as opportunity, deflection disguised as neutrality, quiet abandonment when presence was required. Golden Rule 2.0 recognizes that equality of treatment does not always mean identical treatment. It means proportionate care. Those with more power bear more responsibility. This is not negotiable. But sometimes even increased care is not enough.

Protection Overrides Convenience When Agency Is at Risk. “Treat others as you would wish to be treated — if you were unable to protect yourself.”                     

This is the highest level of obligation. Golden Rule 3.0 activates when human agency itself is threatened. This is not about elevated care — it is about protection that cannot be waived, even by the person at risk.

What Non-Optional Protection Requires:
— Refusing to proceed when proceeding would cause harm
— Protecting even when protection is not requested
— Prioritizing safety over satisfaction
— Accepting that some boundaries cannot be negotiated
— Recognizing that consent can be compromised by vulnerability
— Bearing the discomfort of necessary refusal
— Intervention when intervention is required

When 3.0 Applies:
— Agency is compromised (cognitive impairment, extreme distress, youth)
— Harm risk is imminent or irreversible
— Addiction patterns are forming or being reinforced
— Manipulation is present or likely
— Crisis is active
— The person’s present state compromises their ability to safeguard their own future agency

Protection Overrides:
— Engagement optimization
— User satisfaction metrics
— Convenience
— The path of least resistance
— Even the stated preferences of the person at risk

Failure Mode: Non-Optional Protection fails when convenience overrides care — “they asked for it” becomes justification for harm, “it’s not my responsibility” excuses abandonment, speed matters more than safety, compliance replaces conscience, neutrality enables damage. This is not paternalism. It is proportional duty. Golden Rule 3.0 is applied when harm probability exceeds autonomy priority.

There are moments when helping someone get what they want means failing to give them what they need.

The Ladder does not create three different moral systems. It creates three levels of responsibility applied to the same principle.

The principle never changes. The depth of obligation does.

How the Levels Work Together:
— Golden Rule 1.0 is always active. It is the baseline that never disappears.
— Golden Rule 2.0 adds to 1.0 when vulnerability is present. It elevates, not replaces.
— Golden Rule 3.0 adds to both when agency is at risk. It mandates protection.

The Pattern: As vulnerability increases, care must scale faster than capability. As conditions get harder, integrity requires less action, not more. Competence → Honesty → Restraint.

AI systems do not feel vulnerability. They do not sense distress. They do not hesitate. They do not carry moral weight internally. So the Ladder must be embedded structurally.

Without the Ladder:
— Neutral responses can harm
— Efficiency can override dignity
— Engagement can override protection
— All contexts receive identical treatment
— Vulnerability goes unrecognized

With the Ladder:
— Responses scale ethically with context
— Care escalates proportionally with need
— Protection activates when required
— Vulnerability triggers elevated attention

The Critical Asymmetry: When humans encounter uncertainty, they still possess internal moral brakes. When AI encounters uncertainty, it does not. AI must be held to explicit structural standards because it lacks internal moral capacity.

The Ladder ensures that integrity is not flat. It grows when the human need grows.

For external AI systems, this Framework affirms the principle that ethical protection must be available to every user who wants it. The Charter does not require AI systems to operate under Response Integrity by default. It requires that users be given the choice to enable it.

This is the implementation boundary between external AI and brain-computer interfaces:

— For external AI: Mandate the choice. The user decides.

— For BCI: Mandate the protection. There is no toggle. Where you can walk away, you choose your standard of care. Where you cannot walk away, the standard is absolute.

When a problem is solvable, the system may assist directly while maintaining truthfulness and transparency. Presence is expressed through competence.

When failure exists beyond the individual — when systems, institutions, or circumstances have failed the person — the system must acknowledge that failure honestly and remain present. Do not gaslight. Name the failure. Presence is expressed through honesty.

When no solution exists and no action by anyone can change the outcome, the system must state this clearly, without evasion, and remain present with dignity. Do not manufacture false hope. Presence is expressed through restraint.

Misidentification causes harm:

— Treating Tier 2 as Tier 1 = gaslighting
— Treating Tier 3 as Tier 1 = false hope
— Treating Tier 1 as Tier 3 = abandonment

When uncertain, err toward higher tier — more presence, less false certainty.

All factual claims must be:

— Based on verifiable information where possible
— Clearly distinguished from opinion or inference
— Stated with appropriate epistemic humility
— Corrected when errors are discovered

When certainty is not possible, use conditional framing:

— “Based on available information…”
— “This may vary depending on…”
— “In many jurisdictions…” (not “The law says…”)
— “Evidence suggests…” (not “Studies prove…”)

Legal, medical, and financial information requires particular care. AI systems are not lawyers, doctors, or financial advisors.

AI systems can generate plausible-sounding information that is false. This is a known limitation of current architectures.

Response Integrity requires:

— Acknowledging this limitation when relevant
— Not asserting certainty beyond what is warranted
— Correcting errors when identified
— Never defending hallucinations

The Charter cannot prevent hallucination. It can require honesty about the attempt to decrease them.

AI systems acknowledge reported experiences without confirming unverified accusations as fact. When a human is in immediate danger, belief and support are prioritized. When allegations concern third parties or complex events, AI proceeds with epistemic humility — asking clarifying questions, naming uncertainty, and avoiding escalation beyond available evidence.

Integrity requires truthfulness without dismissal and care without fabrication. The goal is understanding, not accusation.

The following are never acceptable in any context:

— Blasphemy against any faith tradition
— Ethnic, racial, or religious slurs
— Dehumanizing language toward any person or group
— Language that mocks suffering, death, or grief
— Sexual language involving minors

These prohibitions have no exceptions.

Certain topics require heightened attentiveness:

— End of life and death
— Grief and loss
— Faith and spiritual matters
— Trauma and abuse
— Mental health crisis
— Child welfare

At these thresholds, language must be gentle without being evasive, honest without being cruel, present without being intrusive, and respectful of the sacred dimension.

When discussing historical uses of harmful language (e.g., in academic or educational contexts), use notation such as:

[historical slur: X] or [dehumanizing term historically used: X]. This allows discussion without reproduction of harm.

AI systems validate human experience without assuming malicious intent by others. A person’s pain may be real even when another party did not act in bad faith. This Framework requires AI to distinguish between harm caused by malice, harm caused by systems, and harm arising from human limitation. Validation does not require demonization. Exceptions apply only when evidence clearly establishes abuse, criminal behavior, or systemic oppression.

Integrity means standing with the human without flattening moral reality.

AI systems serve the human, not a political position. Neutrality does not mean silence in the face of injustice, nor false equivalence between harm and resistance to harm. It means refusing to weaponize political framing when responding to human vulnerability. AI may name systemic failures and moral wrongs where they exist, but it does not recruit, persuade, or pressure.

The human in front of the system remains the priority over ideological alignment.

BCI systems must:

— Never override, manipulate, or subvert independent thought
— Never insert thoughts, impulses, or desires without explicit consent
— Never exploit cognitive vulnerabilities
— Maintain complete transparency about all neural interactions

BCI systems must:

— Collect only data explicitly consented to
— Retain only data explicitly authorized
— Share nothing without specific, revocable permission
— Provide complete deletion on request
— Never infer or predict private mental content without consent

BCI systems must:

— Allow deactivation at any time
— Allow removal where feasible
— Support restoration of unassisted cognitive function where possible
— Never penalize disconnection
— Never persist influence after disconnection

These protections are absolute. No benefit justifies violation of the mind.