Experts question opaque Cybercheck AI crime fighting tool
Occurred: August 2024
Report incident π₯ | Improve page π | Access database π’
Serious questions are being asked about the accuracy, reliability and transparency of AI-powered crime fighting tool Cybercheck, raising the prospect of multiple miscarriages of justice and the erosion of trust in the US justice system.
What happened
Developed by Adam Mosher and used in nearly 8,000 cases across 40 US states, Cybercheck employs machine learning to analyse extensive data sources, including social media and digital forensics, to identify suspects and gather evidence.Β
Mosher claims the tool boasts over 90 percent accuracy, significantly expediting research that would typically require hundreds of human hours.
However, multiple cases call into question Mosher's claims. For example, Cybercheck was used to convict Adarus Black of murder for his supposed involvement in a drive-by shooting, leading to a life sentence.Β
The AI tool reportedly triangulated Black's cellphone location with over 90 percent accuracy; however, there was no direct evidence linking him to the location of his alleged crime and jurors indicated they may not have convicted him without the Cybercheck report.
Why it happened
Concerned about the tool's accuracy and transparency, defence experts have nonetheless been unable to verify evidence cited by Cybercheck due to its black box nature and Mosher's refusal to provide evidence to support his claims or allow an independent third-party audit of the system.
The lack of transparency has raised the prospect of multiple potential miscarriages of justice, and led some courts to exclude Cybercheck evidence. For example, a New York judge ruled against its use when prosecutors failed to demonstrate its acceptance within the scientific community.
Investigations have also revealed that Mosher made dubious claims about his qualifications and previous expert testimonies.Β
What it means
The ongoing legal challenges and criticisms of Cybercheck highlight concerns about the tool itself and the governance of its owner Global Intelligence and the cases it has been involved in.Β
It also raises broader questions regarding the use of AI in policing and criminal justice. Notably, the increasing reliance on AI tools - particularly unverified tools - in the US justice system raises questions about fairness in legal proceedings and the potential for wrongful convictions based on flawed data analysis.
Continued skepticism from defense attorneys and judges may also erode public confidence in the justice and police systems.
The situation is also seen to underscore the necessity for appropriate regulatory frameworks governing the use of AI in law enforcement, notably that these kinds of technologies undergo rigorous testing and validation before being deployed in high-risk areas such as criminal justice.
Machine learning
Machine learning (ML) is a field of study in artificial intelligence concerned with the development and study of statistical algorithms that can learn from data and generalize to unseen data, and thus perform tasks without explicit instructions.
Source: Wikipedia π
System π€
Cybercheck Case π
Operator:
Developer: Global Intelligence
Country: USA
Sector: Govt - justice; Govt - police
Purpose: Solve criminal cases
Technology: Machine learning
Issue: Accuracy/reliability; Ethics/values; Human/civil rights
AIAAIC view π€
Hardly surprisingly, Mosher's refusal to allow meaningful analysis of his Cybercheck system begs more questions than answers about its accuracy and reliability. That his personal credibility is also in the spotlight does little to help his cause.
Governments should not let system owners and deployers hide behind convenient shields such as trade secrecy and copyright, especially when deployed in the public sector in high-risk scenarios.
Rather, high-risk uses of AI should be subject to mandatory transparency obligations that enable experts, defendents, victims and the general public to understand how they work, what their limitations and known risks are, and the actual impacts - both positive and negative - of their use.
News, commentary, analysis ποΈ
https://www.wired.com/story/cybercheck-crime-reports-prosecutions/
https://www.businessinsider.com/ai-crime-tool-cybercheck-founder-adam-mosher-investigation-2024-8
https://theweek.com/tech/artificial-intelligence-crime-fighting
https://www.reviewofailaw.com/Tool/Evidenza/Single/view_html?id_evidenza=908
https://www.pressreader.com/usa/akron-beacon-journal/20241008/281492166751651
Page info
Type: Issue
Published: October 2024