LivetechWestern cycleAnchors pending

AI-safety research-to-deployment capability-safety-gap?

As of 2 Oct · 23:18 UTCResolves 31 Dec 20280 movers0 pressure moments
Timeline A · Timeline AReality
40%
Crowd 55%Gap 0

By end of 2028, AI safety researchers have stayed ahead of what the newest frontier models can do. New safety techniques like mechanistic interpretability and red-teaming have caught real risks before they escaped the lab. Anthropic, OpenAI, and DeepMind publish findings faster than their own systems improve, and independent auditors can spot dangerous capabilities before deployment happens.

Timeline B · Timeline BReality
60%
Crowd 45%Gap 0

By end of 2028, frontier AI systems deploy faster than safety research can keep up. Labs release new model versions with stronger capabilities while safety papers lag 6-12 months behind. Red-team findings come too late to change deployed systems, and the gap between what we can test in-house and what emerges in the wild has widened.

crowd 55
The quiet paththe one nobody’s watching

Undetected dangerous AI capability deployed to millions forces both timelines into crisis containment rather than measured research pace.

6%
The read

Where reality is heading

A, 80% bandCrowd today
Projected reality support to 31 Dec 2028A rises from 41% to about 52% with an 80% band of 33 to 71. B falls from 53% to about 44%. Quiet path near 4 to 6%. Crowd today: A 55%.100%7550250Crowd today · A 55%A ~52%B ~44%P ~4%NowJan 2027Jan 202831 Dec 2028
Reality crosses toward A in early 2028. Today’s crowd number sits at the top of A’s band, possible but optimistic. View as table

The movers, ranked by leverage

What would change our mind

Signals toward A
    Signals toward B

      Evidence