Incident 28: 2010 Market Flash Crash

Description: A modified algorithm was able to cause dramatic price volatility and disrupted trading in the US stock exchange.


New ReportNew ReportNew ResponseNew ResponseDiscoverDiscover
Alleged: Navinder Sarao , Waddell & Reed and Barclays Capital developed and deployed an AI system, which harmed Market Participants.

Incident Stats

Incident ID
Report Count
Incident Date
Sean McGregor

CSET Taxonomy Classifications

Taxonomy Details

Full Description

On May 6, 2010, the New York Stock Exchange and US Down Jones were greatly impacted by highly volatile trading at high volumes. The blame fell on a single trader in the UK, Navinder Singh Saroa, who allegedly modified a trading algorithm to allow him to mislead the market. Saroa would place requests to purchase stocks (establishing interest in the stock and driving the price higher) but cancel the transaction before it was carried out. Reports say that within minutes major stocks such as General Electric and Accentre had hit $0, and the overall market dropped by 6%. Around $1 trillion in paper stocks had seemingly been wiped out. He faces 22 charges in the US.

Short Description

A modified algorithm was able to cause dramatic price volatility and disrupted trading in the US stock exchange.



Harm Type

Financial harm

AI System Description

A stock trading algorithm designed to quickly detect shifts in stock prices and execute trades accordingly.

Sector of Deployment

Financial and insurance activities

Relevant AI functions

Perception, Cognition, Action

AI Techniques

stock market algorithm, machine learning

AI Applications

stock trading



Named Entities

Navinder Singh Saroa, Dow Jones Industrial Index, Chicago Merchant Exchange

Technology Purveyor

Navinder Singh Sarao

Beginning Date


Ending Date


Near Miss

Harm caused


Deliberate or expected

Lives Lost


Infrastructure Sectors

Financial services

Financial Cost

Short term: $1 trillion unclear of long-term impact

Laws Implicated


Data Inputs

Stock Price, trading volume

GMF Taxonomy Classifications

Taxonomy Details

Known AI Goal

Automatic Stock Trading

Potential AI Technology


Potential AI Technical Failure

Overfitting, Gaming Vulnerability

2010 Flash Crash

2010 Flash Crash

2010 Flash Crash

2010 Flash Crash

The 2010 Flash Crash

The 2010 Flash Crash


A "variant" is an incident that shares the same causative factors, produces similar harms, and involves the same intelligent systems as a known AI incident. Rather than index variants as entirely separate incidents, we list variations of incidents under the first similar incident submitted to the database. Unlike other submission types to the incident database, variants are not required to have reporting in evidence external to the Incident Database. Learn more from the research paper.

Similar Incidents

By textual similarity

Did our AI mess up? Flag the unrelated incidents