Guardrails And Safety Mechanisms For Llm-Powered Enterprise Applications
Keywords:
Large Language Models (LLMs) ,AI Guardrails ,AI Safety ,Enterprise AI ApplicationsAbstract
There is a fast pace of Introducing Big Language Models in Industry Corporate Applications and So a Business Administration is reasonable to use Due to artificial intelligence based even if a Structure and Change of Growth by Making Decisions, the Security of Knowledge can be Constructed, and Decision Making Systems have Occur. But even though LLMs cater to Massive trainability, there are Micro controversies in EIRM because of ARR towards TMS inclusive of but not limited to issues of accountability, security, availability, organizational compliance, privacy breaches, and ethical uses of AI technologies. Scenarios of malfunction of the Models i.e. user input recognized as something it is not or spurious outputs, response modification, leakages and compliance infractions expose corporations to significant threats. To tackle these issues as a result of protective or preventive measures guardrail designs and complementary architectures have gained importance in enterprise level AI systems architecture. This work presents current designs of such guardrail frameworks as well as preventative strategies created for the protection and control of operational l2A applications powered by LLMs with a horizon of 2024. Such safeguards include among others, input and output filtering measures, self-examination mechanisms, compliance with established procedures, as well as responsible behavior modification through human supervision and regulatory enforcement. That is, there are sections addressing policing the behavior of a specified AI and systems policies in AI. The present paper will also address the use of follow-up, compliance, and enforcement operations to sustain safe AI systems. It was impossible to say so, this is ... Hence, as such Executive large scale AI trainer and usage of generative AI has got to proportionate organization instances a better terms.\GeneratedValue general less legal and operational hazards and enhances the enhances the adoption of generative AI in a business setting with LLMs as the primary functional entities. In view of the growing dependence of LLMs for core business operations, there will always be safeguards in place to ensure the operations of AI are safe, secure and compliant.
References
Amodei, D., Olah, C., Steinhardt, J., Christiano, P., Schulman, J., & Mané, D. (2016). Concrete Problems in AI Safety. arXiv. https://arxiv.org/abs/1606.06565
Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., et al. (2021). On the Opportunities and Risks of Foundation Models. Stanford Center for Research on Foundation Models. https://arxiv.org/abs/2108.07258
Floridi, L., Cowls, J., Beltrametti, M., Chatila, R., Chazerand, P., Dignum, V., et al. (2018). AI4People—An Ethical Framework for a Good AI Society. Minds and Machines, 28(4), 689–707. https://doi.org/10.1007/s11023-018-9482-5
Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y., Madotto, A., & Fung, P. (2023). Survey of Hallucination in Natural Language Generation. ACM Computing Surveys, 55(12), 1–38. https://doi.org/10.1145/3571730
Karpukhin, V., Oguz, B., Min, S., Lewis, P., Wu, L., Edunov, S., Chen, D., & Yih, W. T. (2020). Dense Passage Retrieval for Open-Domain Question Answering. Proceedings of EMNLP 2020, 6769–6781.
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W. T., Rocktäschel, T., Riedel, S., & Kiela, D. (2020). Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. Advances in Neural Information Processing Systems, 33, 9459–9474.
NIST. (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0). National Institute of Standards and Technology (NIST)
Perez, F., & Ribeiro, I. (2022). Ignore Previous Prompt: Attack Techniques for Language Models. arXiv. https://arxiv.org/abs/2211.09527
Shneiderman, B. (2022). Human-Centered AI. Oxford University Press.
Weidinger, L., Mellor, J., Rauh, M., Griffin, C., Uesato, J., Huang, P. S., Cheng, M., et al. (2022). Ethical and Social Risks of Harm from Language Models. Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, 615–628. https://doi.org/10.1145/3531146.3533088
Willison, S. (2023). Prompt Injection Attacks Against Large Language Models. arXiv. https://arxiv.org/abs/2302.12173
Downloads
How to Cite
Issue
Section
License
Copyright (c) 2024 International Journal of Engineering Science & Humanities
This work is licensed under a Creative Commons Attribution 4.0 International License.
