普华永道：2023生成人工智能风险管理指南-从治理开始（中译版）（25页）.pdf

1、Managingthe risks ofgenerative AIMay 2023A playbook for risk executives beginning with governanceTable of contents Introduction 2Whats at stake for business?4The new and amplified risks to manage 6For the chief information security officer 8For the chief data officer and chief privacy officer 10For

2、the chief compliance officer 12For the chief legal officer and general counsel 14For internal audit leaders 16For the chief financial officer and controller 18How to get started 20 Bringing it all together 21Bottom line 23Introduction Something truly revolutionary happened in November 2022.Suddenly,

3、anyone with an internet connection,armed only with the ability to hold a conversation in a chat app,could wield the transformative power of artificial intelligence(AI).Within a single week,more than a million users,with ChatGPTs help,produced short-form articles,wrote computer code,made art and summ

4、arized long sources into pieces perhaps better and more concise than the originals.Meanwhile,malicious threat actors test-drove ChatGPT to write malware,more believable phishing emails and more convincing fake identities,rapidly and for widespread dissemination potential harbingers of large-scale fr

5、aud,privacy violations,disinformation and cyber attacks.Mere months after the debut of ChatGPT,generative AI continues to become ever more deeply intertwined into our lives and businesses.Weve seen the fastest ramp-up in active consumer users ever.Weve seen a leap in capabilities from OpenAIs GPT3 t

6、o GPT4 achievements recorded in coding and mid-level professional writing.In quick succession,tech companies have launched/re-launched competing products;start-ups have released models for bespoke applications;and companies,including PwC,have announced massive investments to create their own“Company