Voluntary frameworks (industry codes, technical standards, self-certifications) proportionate to the risk of harm mitigate risk, with low-risk uses needing only basic commitments (transparency reporting, a grievance mechanism) and high-risk uses in health or finance needing more, supported by incentives such as sandbox access, public recognition, directed venture capital and toolkits, and convertible over time into mandatory baselines. AI systems are built to be transparent and fair, open and non-discriminatory, explainable, and secure by design. Techno-legal measures that embed legal requirements in system architecture, such as DEPA for AI training, content authentication, privacy-preserving tools and transparency in consequential automated decisions, buttress policy, regulation and voluntary measures where a clear regulatory objective and broad impact exist, complemented by algorithmic auditing, transparency frameworks and sector rules. Human oversight and other safeguards mitigate loss-of-control risk, with human-in-the-loop at critical decision points so outputs can be reviewed, overridden or supplemented, circuit breakers and system-level constraints where human oversight is too slow, and regular monitoring, testing, audit trails and reporting in critical sectors.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.