Unverified50% confidenceOpinionExact time
AI safety researcher Ajeya Cotra believes that the training infrastructure of next-generation large models may face continuous attacks from thousands of 'radically superhuman'-level AI hackers.
1
Sources
50%
Confidence
Long-term
Relevance
9/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAI training data primarily comes from the internet, so its interpretation of horror tends to lean toward the horror aesthetics prevalent in contemporary online popular culture.71% similarVerifiedCore technical challenges facing AI alignment research include reward hacking, goal generalization failure, deceptive alignment, and scalable oversight.70% similarUnverifiedThe proliferation of generative AI has lowered the barrier to cyberattacks: large language models can generate highly customized phishing emails within seconds, and AI tools can assist in analyzing open-source code vulnerabilities and automatically generating exploitable attack code70% similarUnverifiedOnce sensitive information enters the AI context, there are multiple potential risk surfaces: log retention, mixing into model training data, and possible exfiltration behavior if the AI is granted network access.70% similarVerifiedA single complete task execution by an AI coding agent may involve consumption of tens of thousands to even hundreds of thousands of tokens.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/891403API
curl https://kongchang.com/api/v1/knowledge/claims/891403MCP
get_claim(id=891403)