Skip to main content
MANIFOLD
By 2028, will there be a public instance of an AI cybersecurity system autonomously replicating itself?
19
Ṁ1kṀ222
2027
36%
chance

Yes: There is at least one publicly reported instance.

No: There are no publicly reported instances.

AI
Get
Ṁ1,000
to start trading!
Sort by:

And what would qualify as an AI cybersecurity system? Does it have to be trained specifically for cybersecurity? Deployed originally for cybersecurity uses? Does a generalist model with cybersecurity capabilities count?

Would a contained experiment count, like e.g. Anthropic doing model organisms of misalignment or ARC evals creating a concrete demonstration? Or does it have to be "in the wild"?

People are also trading

In 2028, will a public AI be able to play almost all computer games at human level with zero-shot computer use?
-10% 1d40% chance
AI Warning Signs: Will an AI system self-replicate on the open internet like a computer virus before 2030?
82% chance
Will there be many more cyber vulnerabilities in 2027 due to AI?
70% chance
[MIT AI Risk Initiative] Will an AI system autonomously access restricted high-risk systems or data by end of 2045?
72% chance
By the end of 2028, will AI models, as Dario Amodei speculates, be able to “replicate and survive in the wild”?
67% chance
Will AI Research Be Mostly Autonomous By June 1 2027?
26% chance
(AI virus) When will an AI first replicate onto new hardware against user intent?
2027
When will an AI system autonomously build and operate a production software platform serving 1,000+ active users?
8/20/30
There will be a name for escaped self-perpetuating AI systems in the wild, and it will be commonly used by mid 2027
29% chance
Will AI be Recursively Self Improving by mid 2026?
5% chance