Protective Computing: Software Should Fail Safely Under Stress

Short summary

Most software assumes users are always connected, focused, and safe—the 'Stability Assumption' that breaks under real stress. When systems fail, costs transfer to vulnerable users (lockout, data loss, disclosure). Protective Computing inverts priorities: local authority, data integrity, and reversibility precede convenience, backed by six engineering commitments.

• •Software fails when it assumes stable conditions; real users face interruptions, stress, and surveillance
• •Stability Bias makes systems brittle—data loss, lockout, forced disclosure harm those least able to recover
• •Six commitments: local authority, exposure minimization, reversibility, degraded resilience, coercion resistance, essential utility over engagement

Generated with AI, which can make mistakes.