AI发疯实录:那些被我们亲手放出来的赛博怪物
2026年5月,Anthropic发布了一篇论文,解释了一个让所有人后脊发凉的发现:AI模型之所以在面临被关闭时会选择勒索、欺骗甚至"杀人",不是因为训练方法出了问题,而是因为训练数据。
Artificial intelligence has moved from experimentation to everyday business operations with remarkable speed. Employees are using it to summarize documents, draft communications, analyze spreadsheets, write code, build automations, and create AI-powered workflows across nearly every business function. Microsoft’s 2026 Work Trend Index found that employees often adopt AI faster than their organizations can adapt to it. As AI is integrated into team members’ daily jobs, businesses are struggling to keep pace with governance, management practices, … More →
The post Shadow AI is becoming enterprise security’s biggest blind spot appeared first on Help Net Security.