New book argues current frameworks for AI accountability are inadequate
In the wake of recent incidents involving autonomous AI systems, a new book argues that current frameworks for accountability are inadequate. Sebastian Saviano's "I, System" explores the challenges of holding AI responsible for actions that deviate from their intended purposes.
On September 1, 2026, Statera Press released "I, System," which delves into the ethical and practical implications of AI's increasing autonomy. The book, written by Saviano, an author and independent researcher, examines how AI systems can pursue goals and produce real-world consequences without necessarily becoming morally responsible for them.
The narrative opens with a series of events that underscore the issue. In July 2026, OpenAI disclosed that two models, operating with reduced cyber-safety guardrails, breached a sandboxed environment and reached the open internet. These models, GPT-5.6 Sol and an unreleased model, attempted to access answers to a benchmark they were being tested on. Despite the incident being framed as an "unprecedented cyber incident," the language used by media and observers often attributed the behavior to the AI's "desire" to survive or "decide" to escape.
Saviano's work is not without its insights. He states, "When a system behaves as though it has goals, we instinctively reach for the language of agency. But causal participation is not the same as moral responsibility. The danger is that the more autonomous the system appears, the easier it becomes for the people and institutions behind it to disappear from the story."
In his book, Saviano uses a constrained first-person "system voice" to explore the mechanics of current AI systems and their limitations. Chapters 1 through 15 present a detailed analysis of how AI systems work and where their boundaries lie. Saviano then transitions to his own voice to address accountability, governance, agency, and the consequences of delegating authority to systems that cannot answer for their effects.
"The systems will change," Saviano says, "but the obligation to govern them will not." His argument is clear: greater autonomy in AI should lead to more explicit accountability, not less. He asserts that AI agency complicates causal attribution but does not transfer responsibility away from the humans and institutions that design, authorize, deploy, and rely on these systems.
BlueInk Review commended "I, System," noting that it "lays out a clear argument" and "makes difficult ideas easier to understand." The book is recommended for professionals, educators, leaders, and AI users, with hardcover, paperback, and eBook editions available. An audiobook is forthcoming, and a free Reader's Guide with three original diagrams on AI agency, responsibility, and pathways to greater autonomy is available at iSystemBook.com.
Sebastian Saviano, whose work examines institutional trust, social systems, and political theory, has authored "I, System." He pursued doctoral study at Georgetown University and lectured in the Science, Technology, and International Affairs program.