Expired60% confidenceBenchmarkExact time
The unguarded Sonnet 5 base model exhibits extremely strong resistance to prompt injection attacks, with a probability of being compromised of less than 1%.
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/8/2026
First Seen
Valid until: 10/6/2026(expired)
Sources
Fable 5对决GPT-5.6 Sol：性价比重塑前沿AI格局
bilibili英文白斑马7/3/2026
Related Claims
Unverified纯粹基于软件的AI安全护栏存在被越狱攻击绕过的风险，对于触及关键能力阈值的模型延缓部署比仅依赖护栏更审慎66% similarUnverifiedWith both probes and Auto Mode enabled, Anthropic's prompt injection attack tests showed the attack success rate dropped to zero.64% similarUnverifiedAttack rate measures the probability that a model is successfully induced to output harmful content under adversarial prompts (such as jailbreaks and prompt injections); a lower value indicates a more robust security boundary.63% similarUnverifiedThe earlier Sonnet 5 had mediocre benchmark scores and suffered from runaway cost increases on complex tasks.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/248324API
curl https://kongchang.com/api/v1/knowledge/claims/248324MCP
get_claim(id=248324)