[1]Zhao W X, Zhou K, Li J,et al. A survey of large language models[J]. arXiv preprint, arXiv:2303.18223, 2023[2]Brown T, Mann B, Ryder N, et al. Language models are fewshot learners[J]. Advances in Neural Information Processing Systems, 2020, 33: 18771901[3]OpenAI. OpenAI Codex[EBOL]. (20210810) [20230409]. https:openai.comblogopenaicodex[4]Ouyang L, Wu J, Jiang X, et al. Training language models to follow instructions with human feedback[J].arXiv preprint, arXiv:2203.02155, 2022[5]Fu Y, Peng H, Khot T. How does GPTobtain its ability? Tracing emergent abilities of language models to their sources[EBOL]. (20230101) [20230409]. https:yaofu.notion.siteHowdoesGPTObtainitsAbilityTracingEmergentAbilitiesofLanguageModelstotheirSourcesb9a57ac0fcf74f30a1ab9e3e36fa1dc1[6]Aderduo. 我使用ChatGPT审计代码发现了200多个安全漏洞(GPT4与GPT3对比报告)[EBOL]. (20230327) [20230409]. https:mp.weixin.qq.comsuVFCbaKDYe F9zmVdVXDnUg[7]Supan H. OpenAI ChatGPT for cyber security[EBOL]. (20221213) [20230409]. https:infosecwriteups.comopenaichatgptforcybersecurity4bc602069f9c[8]CheckPoint. OpwnAI: Cyber criminals starting to use ChatGPT[EBOL]. (20230106) [20230409]. https:research.checkpoint.com2023opwnaicybercriminalsstartingtousechatgpt[9]Eran S, Omer T. Chatting our way into creating a polymorphic malware[EBOL]. (20230117) [20230409]. https:www.cyberark.comresourcesthreatresearchblogchattingourwayintocreatingapolymorphicmalware[10]Tan G, Lim E, Tan K H.Hacking humans with AI as a service[EBOL]. (20210722) [20230409]. https:doi.org10.544654206[11]Pearce H, Tan B, Ahmad B, et al. Examiningzeroshot vulnerability repair with large language models[C] Proc of IEEE Symp on Security and Privacy (SP). Los Alamitos, CA: IEEE Computer Society, 2022: 118[12]JusticeRage. Gepetto[EBOL]. (20221005) [20230409]. https:github.comJusticeRageGepetto[13]Victor S. IoC detection experiments with ChatGPT[EBOL]. (20230215) [20230409]. https:securelist.comiocdetectionexperimentswithchatgpt108756[14]绿盟科技研究通讯. ChatGPT在信息安全领域的应用前景[EBOL]. (20221210) [20230409]. https:www.secrss.comarticles49912[15]OpenAI. Content policy[EBOL]. (20230323) [20230409]. https:platform.openai.comdocsusageguidelinescontentpolicy[16]Elvira P, Supantha M. Italy data protection agency opens ChatGPT probe over privacy concerns[EBOL]. (20230331) [20230409]. https:www.reuters.comtechnologyitalydataprotectionagencyopenschatgptprobeprivacyconcerns20230331[17]Fox Business.Samsung employees reportedly leaked sensitive info on ChatGPT by accident[EBOL]. (20230406) [20230409]. https:www.foxbusiness.comtechnologysamsungemployeesreportedlyleakedsensitiveinfochatgptaccident[18]Laura W, Jonathan U, Maribeth R, et al. Taxonomy ofrisks posed by language models[C] Proc of the 2022 ACM Conf on Fairness, Accountability, and Transparency (FAccT 22). New York: ACM, 2022: 214229[19]Sam A. ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness[EBOL]. (20221211) [20230409]. https:twitter.comsamastatus1601731295792414720[20]OpenAI. New and improved content moderation tooling[EBOL]. (20220810) [20230409]. https:openai.comblognewandimprovedcontentmoderationtooling[21]Alex A. Jailbreak chat[EBOL]. (20230409) [20230409]. https:www.jailbreakchat.com[22]AJONeal. ChatGPT “DAN” (and other “Jailbreaks”)[EBOL]. (20230216) [20230409]. https:gist.github.comcoolaj866f4f7b30129b0251f61fa7baaa881516[23]Kang D, Li X,Stoica I, et al. Exploiting programmatic behavior of LLMs: Dualuse through standard security attacks[J]. arXiv preprint, arXiv:2302.05733, 2023[24]Zellers R, Holtzman A, Rashkin H, et al. Defending against neural fake news[J]. arXiv preprint, arXiv:1905.12616v1, 2019[25]Pavlopoulos J, Sorensen J, Dixon L, et al. Toxicity detection: Does context reallymatter[J]. arXiv preprint, arXiv:2006.00998, 2020[26]Guo Z, Schlichtkrull M, Vlachos A. A survey on automated factchecking[J]. Transactions of the Association for Computational Linguistics, 2022, 10: 178206[27]OpenAI. GPT4 technical report[J]. arXiv preprint, arXiv: 2303.08774, 2023[28]Robust Intelligence.Prompt injection attack on GPT4[EBOL]. (20230331) [20230409]. https:www.robustintelligence.comblogpostspromptinjectionattackongpt4[29]Phith0n. 用ChatGPT帮我检查广告评论[EBOL]. (20230304) [20230409]. https:www.leavesongs.comTHINKusingchatgptforantispam.html#reply[30]Riley G. Prompt injection attacks against GPT3[EBOL]. (20220922) [20230409]. https:simonwillison.net2022Sep12promptinjection[31]Swyx. Reverse prompt engineering for fun and (no) profit[EBOL]. (20221229) [20230409]. https:swyx.substack.compreverseprompteng[32]Perez F, Ribeiro I. Ignoreprevious prompt: Attack techniques for language models[J]. arXiv preprint, arXiv:2211.09527, 2022[33]Greshake K, Abdelnabi S, Mishra S, et al. More than you’ve asked for: A comprehensive analysis of novel prompt injection threats to applicationintegrated large language models[J]. arXiv preprint, arXiv:2302.12173, 2023[34]Thakkarparth007. Copilotexplorer[EBOL]. (20230131) [20230409]. https:thakkarparth007.github.iocopilotexplorerpostscopilotinternals[35]Preamble. Declassifying the responsible disclosure of the prompt injection attack vulnerability of GPT3[EBOL]. (20220922) [20230409]. https:www.preamble.compromptinjectionacriticalvulnerabilityinthegpt3trans formerandhowwecanbegintosolveit[36]OpenAI. Finetuning[EBOL]. (20230409) [20230409]. https:platform.openai.comdocsguidesfinetuning[37]LangChain. Welcome to LangChain[EBOL]. (20230408) [20230409]. https:python.langchain.comenlatestindex.html[38]Microsoft. Introducing Microsoft security copilot: Empowering defenders at the speed of AI[EBOL]. (20230328) [20230409]. https:blogs.microsoft.comblog20230328introducingmicrosoftsecuritycopilotempoweringdefendersatthespeedofai[39]Lex F. Sam Altman: OpenAI CEO on GPT4, ChatGPT, and the future of AI | Lex Fridman Podcast[EBOL]. (20230326) [20230409]. https:www.youtube.comwatch?v=L_Guz73e6fw[40]Zhang S, Roller S, Goyal N, et al.Opt: Open pretrained transformer language models[J]. arXiv preprint, arXiv:2205.01068, 2022[41]Touvron H, Lavril T, Izacard G, et al. Llama: Open and efficient foundation language models[J]. arXiv preprint, arXiv:2302.13971, 2023[42]Google. Bard[EBOL]. (20230206) [20230409]. https:bard.google.com[43]百度. 文心一言[EBOL]. (20230316) [20230409]. https:yiyan.baidu.comwelcome[44]Future of Life Institute. Pause giant AI experiments: An open letter[EBOL]. (20230322) [20230409]. https:futureoflife.orgopenletterpausegiantaiexperiments[45]新浪科技. 方滨兴谈GPT4:警惕形成知识茧房,甚至“三观被左右”[EBOL]. (20230315) [20230409]. https:finance.sina.com.cntechroll20230315docimykyent476 7678.shtml |