THE DEFINITIVE BLUEPRINT FOR GOVERNING ARTIFICIAL INTELLIGENCE — AND THE ONE POLICYMAKERS CAN ACTUALLY USE
In January 2021, a national government resigned. Not over a war or a budget — over a risk-scoring model. Tens of thousands of families had been wrongly branded as fraudsters by a system nobody voted for, in a wealthy country with strong courts, a free press, and world-leading data protection law. Its institutions took seven years to see what a moderately complex piece of software was doing inside its own tax office.
That is the governance gap. This book closes it.
THE ALGORITHMIC STATECRAFT is the most comprehensive blueprint yet written for governing artificial intelligence — 34 chapters across seven parts, built not on principles and warnings, but on machinery that works.
WHY THIS BOOK IS DIFFERENT
Most AI books explain the technology or sound the alarm. This one hands you the instruments.
At its centre is the Seven-Layer Model of AI Governance — constitutional, statutory, institutional, technical, organisational, deployment, and redress — with a maturity self-assessment any ministry can score itself against in a single week. Around it sit more than thirty original frameworks: the Translation Ladder that converts a principle into an enforceable obligation; the Two-Axis Risk Matrix that classifies systems without freezing innovation; the Four Conditions of Meaningful Human Involvement; the Coercion Test for policing and the courts; the Capacity Ladder, costed at three jurisdiction sizes; the Interoperability Protocol; and the Compute Access Compact.
INSIDE
- Why the state is not merely late but mis-shaped — and the four gaps to close
- Compute as territory: the geopolitics of chips, cloud, energy and water
- Brussels, Washington, Beijing and the Unaligned Majority — four rival models, compared without a hero
- Human dignity made operational: three testable requirements, with model statutory text
- Why fairness is not a property a system can have, but a decision about who absorbs the errors
- What post-hoc explanation actually delivers — and why regulation built on it will fail
- Liability, redress, and the evidential wall that defeats claimants before doctrine is reached
- AI inside government: the failures, the doctrines that already prohibit them, and a ten-duty operating standard
- Lethal autonomy, cyber operations, biosecurity, and governing systems we do not yet understand
- The international settlement: treaty design, mutual recognition, verification, and enforcement without empire
- A ten-year implementation agenda with owners, dates, indicators — and a ninety-day start of which eleven items require no legislation at all
WHO IT IS FOR
Ministers and officials who must act before consensus arrives. Regulators building capability from nothing. Diplomats preparing negotiating mandates. Executives mapping the compliance world ahead. Lawyers, researchers and journalists who need the field's arguments in one place. And any citizen who wants to know what to ask — and of whom.
Rigorous enough for a treaty office. Readable enough for a flight.
ABOUT THE AUTHOR
Abhijeet Sarkar is CEO and Founder of Synaptic AI Lab Private Limited. He writes at the intersection of frontier technology and institutional design, for readers who must decide rather than speculate.
The question was never whether artificial intelligence would be governed. It is already governing. The question is whether it will be governed by anyone we chose.
