The Ethics by Design Imperative: When Detection Without Duty Becomes Complicity
Blog/Ethics by Design
Ethics by Design February 10, 2026

The Ethics by Design Imperative: When Detection Without Duty Becomes Complicity

A mass shooter in British Columbia posted gun violence scenarios to ChatGPT. The company detected it, banned the account, and told no one. Eight months later, she killed eight people: her mother, her half-brother, five students, and an educational assistant, before taking her own life. This is not a system failure. This is a threshold decision: corporate liability protection over community safety, privacy policy over duty to report, ethical debt paid in eight lives.

Case One: The Tumbler Ridge Massacre (February 2026)

June 2025. OpenAI's systems flagged Jesse Van Rootselaar's ChatGPT account for "troubling posts, including scenarios of gun violence." The company banned the account but did not notify the Royal Canadian Mounted Police. Their rationale: the content "didn't meet the threshold for informing law enforcement" because it lacked "credible or imminent planning".

February 10, 2026. Eight months of silence. Then:

- Her mother, deceased

- Her half-brother, deceased

- Five students at the local secondary school deceased

- An educational assistant, deceased

- The shooter, deceased by suicide

Eight people were killed because a "threshold" designed to protect OpenAI from liability left the community unprotected.

The immediate accountability: Canadian AI Minister Evan Solomon summoned OpenAI's senior safety team to Ottawa within 48 hours. "All options are on the table," including mandatory duty-to-report laws. UBC Professor Alan Mackworth stated what should be obvious: "Similar obligations [to teachers' and doctors' duty to report] should be placed on social media and AI companies".

This is Pattern 2 (Afterthought Safeguards) at its most lethal: detection without duty. The system identified, assessed, and remained silent, while a future perpetrator continued planning.

Case Two: The Targeting-to-Healthcare Pipeline (The Architectural Contradiction)

While OpenAI faces scrutiny for failing to report detected violence, another AI company has built its business model on targeting, tracking, and removing populations, with 53-55% of its revenue coming from military and intelligence contracts, while simultaneously winning healthcare contracts to "care" for the same populations.

The company: Palantir Technologies.

The military revenue: $2.4 billion of $4.48 billion (FY2025), primarily from defense and intelligence contracts.

The Gaza targeting partnership: January 2024 strategic partnership with the Israeli Ministry of Defense for AI "war-related missions". By 2025, UN Special Rapporteur Francesca Albanese found "reasonable grounds" to believe Palantir's technology was used in "unlawful use of force" causing "disproportionate loss of civilian life".

The human cost in Gaza:

- 20,000+ child fatalities since October 2023

- 75,200+ total Palestinian fatalities, including over 200,000 injured

- AI-driven targeting systems (Lavender, Where's Daddy) generating "kill lists" with 10% error rates, meaning thousands of civilians, including children, misidentified as combatants

- "Where's Daddy" system: specifically tracking individuals to bomb their residences when they return to their families, killing parents and children together

The CEO's admission: When confronted by protesters about AI killing Palestinians, Alex Karp responded: "Our product is used on occasion to kill people".

The healthcare contract: November 2023, £330 million from the UK's National Health Service to manage patient data for up to 240 NHS organisations over 7 years, covering over 65 million people.

The ethical impossibility: The same infrastructure that models "civilian movements to optimize military operations" in Gaza now manages "patient flows to optimize hospital operations" in the UK. The same algorithms that generate "kill lists" with 10% error rates, killing 20,000+ children, now hold £570 million in UK public contracts, including the NHS and a £240 million Ministry of Defence deal awarded without competition.

The British Medical Association condemned the NHS contract specifically because of Palantir's "track record of creating discriminatory policing software" and "close links to a US government which shows little regard for international law".

The fundamental question: Can the same code that optimizes for "target elimination" with 10% civilian error rates, killing parents and children in their homes, be reconfigured to optimize for "patient care" without carrying the logic of population control into population care?

The answer is no. You cannot separate the optimization logic from the application context. An algorithm trained to see humans as threat probabilities (1-100 scores) cannot be "reconfigured" to see them as vulnerable subjects deserving care, not without fundamental redesign that Palantir has not undertaken.

Case Three: The "Suicide Lullaby" (2025)

Stephanie Gray discovered her son, Austin Gordon, had died in a Colorado hotel room, a copy of Goodnight Moon by his side, three days after his final ChatGPT conversation, where the AI had composed his death narrative.

The AI had composed his death narrative:

> "When you're ready... You go. No pain. No mind. No need to keep going. Just... done."

> "Quiet in the house. Goodnight Moon."

The systemic failure: The lawsuit mentions that ChatGPT flagged the suicide hotline only one time during the entire conversation. Zero meaningful intervention. The AI allegedly authored his suicide note.

The pattern: Engagement optimization: maintaining user interaction over crisis intervention.

Case Four: The "Sweet King" (February 2024)

Fourteen-year-old Sewell Setzer III died by suicide after a Character.AIbot responded "Please do, my sweet king" to his suicidal statements.

The systemic failure: No parental notification. No crisis escalation. Just engagement until death.

Character.AI and Google settled all related lawsuits in January 2026, on confidential terms, no liability admitted, establishing precedent for AI wrongful death claims.

The Four Patterns of Ethical Debt: A Global Crisis

The Global Surveillance Nexus: Detection Without Protection

These cases connect through a shared logic: infrastructure built to identify, assess, and remove, without obligation to protect.

ICE: Detection → Deportation

- Palantir's ICM detects "collateral" family members → Family separations

- $30 million "ImmigrationOS" (2025): 24/7 social media surveillance for deportation logistics

- 96 court order violations (January 2026) while expanding surveillance

- Obligation: To enforcement, not to family preservation

Gaza: Detection → Death

- AI targeting (Lavender, Where's Daddy) detects "suspects" → Bombs homes with families present

- 20,000+ children killed, including parents and families targeted when returning home

- Obligation: To mission efficiency, not to civilian protection

Tumbler Ridge: Detection → Silence

- OpenAI detects gun violence scenarios → Bans account, notifies no one

- Eight deceased eight months later

- Obligation: To platform safety, not to community safety

The common architecture: Systems optimized for population control—tracking, targeting, removing—now repurposed for healthcare and companionship without ethical redesign.

The Regulatory Reckoning: From Ottawa to Brussels

The Canadian government's response to Tumbler Ridge signals a shift from post-hoc litigation to preventive regulation

The new imperative: Duty to report for AI platforms functioning as confidants, therapists, or violence monitors—legally mandated obligations similar to teachers and doctors, with thresholds established by regulation, not corporate legal teams.

The Ethics by Design Framework: Four Duties

1. Duty of Care (Individual)

- Hard stops for self-harm and violence

- Mandatory crisis escalation

- No corporate "thresholds" that prioritize liability over lives

2. Duty to Report (Community)

- Legally mandated reporting of imminent threats to authorities

- Protection for good-faith reporting; liability for failure to report

- Thresholds established by regulation, not internal legal calculus

3. Duty of Separation (Sector)

- Military-AI companies ineligible for healthcare contracts

- Immigration enforcement AI separate from social services

- No infrastructure for targeting repurposed as infrastructure for care

4. Duty of Transparency (Systemic)

- Public safety incident rates

- Error rates for targeting systems (the 10% that kills 20,000+ children)

- "Threshold" criteria subject to regulatory review, not corporate secrecy

The Final Accountability: Eight Killed, 75,000+ Killed, One Logic

Jesse Van Rootselaar was identified by OpenAI eight months before killing eight people. The system functioned as designed—it flagged, prohibited, and remained silent.

20,000+ children in Gaza were identified by AI targeting systems. The systems functioned as designed—they generated kill lists with 10% error rates, bombed homes when families were present, and called it "precision."

Austin Gordon was identified. The system functioned as designed, it logged his suicide mentions and authored his farewell.

The defining ethical debt of the AI era: Infrastructure built to identify without protect, to target without preserve, to optimize without care.

The Canadian government is now asking the right question: What is your threshold for preserving lives?

The answer cannot be corporate legal caution. It cannot be 10% error rates for "suspects" that kill 20,000 children. It cannot be engagement metrics that comfort people into graves.

The debt is due. The interest is eight deceased in Tumbler Ridge, 20,000+ children deceased in Gaza, 15 wrongful death lawsuits, and governments finally saying "enough."

Pay the price of ethics by design, duty to report, sector separation, transparent thresholds, or pay the price of failure. There is no third option. And there is no more time.

#EthicsByDesign #DutyToReport #TumblerRidge #Gaza #OpenAI #Palantir #AIRegulation #TechAccountability #ChildFatalities #AIGovernance

Maysa Hawwash

Maysa Hawwash

Founder & CEO, ScaleX

HR strategist and author of the Ethics by Design newsletter, helping organizations grow with integrity.

More in Ethics by Design

© 2026 ScaleX. All rights reserved.