Public records show C3 AI's Project Sherlock, its flagship contract to speed up policing in San Mateo County, has struggled with usability and other issues
San Mateo County pumped $12 million into a grand AI surveillance system made by billionaire Tom Siebel's C3 AI that was supposed to level up its police work. LinkedIn: Thomas Brewster . Bluesky: @hypervisible.blacksky.app and @tprstly LinkedIn: Thomas Brewster : 🚨 EXCLUSIVE 🚨 Billionaire Tom Siebel and his company C3 AI, in partnership with Amazon, set up a sprawling $12 million AI surveillance system in San Mateo County. … Bluesky: @hypervisible.blacksky.app : Real shame when your panopticon doesn't work out as planned. Theo Priestley / @tprstly : Time and again US police is conned into spending millions on an AI that is supposed to “predict crime”, and time and again it fails to deliver any expected results. [embedded post]
Context & Ripple Effects
San Mateo County committed $12 million to C3 AI’s Project Sherlock as a flagship effort to improve police work. The reported usability problems turn that deployment into a concrete test of whether a large, integrated surveillance system can work in frontline public-sector operations.
The report also lands against prior accounts of product delays and internal issues at C3.ai and broader legal concerns over the reliability of AI-generated investigative evidence. That makes operational performance—not merely system capability—the central question.
First-order effects
- San Mateo County police users and administrators must contend with a system that has not performed as expected, potentially limiting the value they can extract from the $12 million deployment.
- C3 AI’s flagship public-safety reference project faces added credibility pressure, while its Amazon partnership is tied to a deployment whose usability is under scrutiny.
Second-order effects
- County procurement and oversight teams are likely to place greater weight on demonstrated usability, implementation support, and auditability when evaluating comparable police AI systems.
- Vendors selling investigative and surveillance tools face a sharper burden to distinguish operationally dependable products from systems whose outputs may be difficult for officers and legal stakeholders to rely on.
Third-order effects
- If similar deployments repeatedly falter, public-safety AI purchasing could shift from broad platform promises toward narrower, independently testable tools with clearer accountability for deployment outcomes.
- The pattern strengthens the case for public-safety AI governance that evaluates not only surveillance reach, but also whether systems are usable, reliable, and defensible in real investigative workflows.
The trend: Project Sherlock is one data point in the shift from ambitious AI-policing deployments toward closer scrutiny of whether surveillance systems deliver dependable public-sector operations.