Accountability in Agentic Secure Software Development Life Cycles: Challenges, Degradation Scenarios, and Governance Principles

• Rubens Abraão da Silva Sousa UECE
• Gabriel Pinheiro Rezende de Lima UECE
• Igor Ramsés Temóteo dos Santos UECE
• Sávio Freire IFCE / UECE
• Ismayle de Sousa Santos UECE
• Paulo Henrique Mendes Maia UECE

DOI: https://doi.org/10.5753/se4as.2026.33008

Resumo

Secure software development frameworks, such as OWASP SAMM, BSIMM, and NIST SSDF, structure security decisions on the premise that an identifiable professional approves, audits, and is accountable for them. LLM-based agents disrupt this premise: in threat modeling, penetration testing, and incident response, they not only produce evidence but also propose strategies, prioritize actions, and execute workflows with varying autonomy. Formal responsibility remains human, while the construction of the decision is distributed between humans and agents. This position paper characterizes this accountability decoupling and its resulting phenomenon, the Accountability Chasm: the gap between the operational autonomy of agents and the capacity of governance mechanisms to preserve traceability, contestability, and attribution over security decisions. Unlike the classic accountability gap, here, accountability remains attributed to a human; what degrades is what makes this attribution effective. This paper describes three degradation scenarios in real-world security activities, propose four accountability-oriented governance principles, and presents a research agenda.

Palavras-chave: Accountability, Agentic AI, AI Agents, Secure Software Development Lifecycle, Software Security, AI Governance, Human-Agent Collaboration

Referências

Aldeida Aleti, Baishakhi Ray, Rashina Hoda, and Simin Chen. 2026. Trustworthy AI Software Engineers. arXiv:2602.06310 [cs.SE] [link]

Gauhar Ali, Sajid Shah, andMohammed ElAffendi. 2025. Enhancing cybersecurity incident response: AI-driven optimization for strengthened advanced persistent threat detection. Results in Engineering 25 (3 2025), 104078. DOI: 10.1016/j.rineng.2025.104078

Eranga Bandara, Amin Hass, Sachin Shetty, Ravi Mukkamala, Ross Gore, Abdul Rahman, and Safdar H. Bouk. 2025. Deep-Stride: Automated Security Threat Modeling with Vision-Language Models. In 2025 International Conference on Software, Telecommunications and Computer Networks (SoftCOM). IEEE, 1–7. DOI: 10.23919/SoftCOM66362.2025.11197424

Alan Chan, Carson Ezell, Max Kaufmann, Kevin Wei, Lewis Hammond, Herbie Bradley, EmmaBluemke, Nitarshan Rajkumar, David Krueger, NoamKolt, Lennart Heim, and Markus Anderljung. 2024. Visibility into AI Agents. In The 2024 ACM Conference on Fairness, Accountability, and Transparency. ACM, Rio de Janeiro Brazil, 958–973. DOI: 10.1145/3630106.3658948

Ahmed E. Hassan, Hao Li, Dayi Lin, Bram Adams, Tse-Hsun Chen, Yutaro Kashiwa, and Dong Qiu. 2025. Agentic Software Engineering: Foundational Pillars and a Research Roadmap. (9 2025). [link]

JundaHe, Christoph Treude, andDavid Lo. 2025. LLM-BasedMulti-Agent Systems for Software Engineering: Literature Review, Vision, and the Road Ahead. ACM Transactions on Software Engineering and Methodology 34 (6 2025), 1–30. Issue 5. DOI: 10.1145/3712003

Frank Hindriks and Herman Veluwenkamp. 2023. The risks of autonomous machines: from responsibility gaps to control gaps. Synthese 201, 1 (Jan. 2023), 21. DOI: 10.1007/s11229-022-04001-5

Rashina Hoda. 2026. Toward Agentic Software Engineering Beyond Code: Framing Vision, Values, and Vocabulary. (2 2026). [link]

Rafiq Ahmad Khan, Siffat Ullah Khan, Muhammad Azeem Akbar, and Musaad Alzahrani. 2024. Security risks of global software development life cycle: Industry practitioner’s perspective. Journal of Software: Evolution and Process 36 (3 2024). Issue 3. DOI: 10.1002/smr.2521

Andreas Matthias. 2004. The responsibility gap: Ascribing responsibility for the actions of learning automata. Ethics and Information Technology 6, 3 (2004), 175–183. DOI: 10.1007/s10676-004-3422-1

Gary McGraw and Brian Chess. 2009. The Building Security in Maturity Model (BSIMM). USENIX Association, Montreal, Quebec.

National Institute of Standards and Technology. 2026. Secure Software Development Framework SSDF. [link]. Acesso em: 16 maio 2026.

OWASP Foundation. n.d.. OWASP SAMM. [link]. Acesso em: 16 maio 2026.

Filippo Santoni De Sio and Giulio Mecacci. 2021. Four Responsibility Gaps with Artificial Intelligence: Why they Matter and How to Address them. Philosophy & Technology 34, 4 (Dec. 2021), 1057–1084. DOI: 10.1007/s13347-021-00450-x

Xiangmin Shen, Lingzhi Wang, Zhenyuan Li, Yan Chen, Wencheng Zhao, Dawei Sun, Jiashui Wang, and Wei Ruan. 2025. PentestAgent: Incorporating LLM Agents to Automated Penetration Testing. In Proceedings of the 20th ACM Asia Conference on Computer and Communications Security. ACM, Hanoi Vietnam, 375–391. DOI: 10.1145/3708821.3733882