AI Safety & Governance · Projected milestone
Personal AI screening guardian
65% by 2035 · including earlier years
The year is an editorial timing estimate. The percentage applies to occurrence by the stated deadline. These estimates are not calibrated odds.
A user-controlled AI screens communications and agent requests for scams and manipulation, with visible limits.
Explore this milestone on the interactive mapWhat the forecast means
Milestone stage: deployment
partially-demonstrated
Contextual assistants and security controls exist; integrated tested cross-channel guardians are incomplete.
Completion criteria: A broadly purchasable service or device covers at least three communication channels, publishes prospective fraud/manipulation detection results and false positives, lets users inspect/reverse decisions, and operates without requiring expert-level general AI.
This page describes a forecast, rather than an achieved result. Read the methodology and assumptions.
Built on
Led to
Sources & evidence
- Siri AI and the next generation of Apple Intelligence 2026-09-14
Apple announces beta rollout of richer personal context and systemwide app actions, plus on-device models for some features and cloud models for others. This is not offline expert-level general AI or access to every app/service.
- Building commerce agents with Claude 2026-09-02
Anthropic provides commerce-agent reference implementations and describes deployed partner workflows with permission and payment integration. These support narrower useful agents without needing general AI.
- Investigating unintended model actions 2026-10-09
Anthropic reports unintended actions in evaluation/internal settings and remedial controls. This is evidence of a control/reliability gap, not a measurement of the absence of underlying cognitive capability.
Record reviewed 2026-10-18