Research On Criminal Accountability For Autonomous Financial Trading Errors
1. Michael Coscia – Spoofing with Algorithms (U.S., 2015)
Facts:
Coscia used high-frequency trading algorithms to place orders he intended to cancel immediately (spoofing).
He placed large orders to create the illusion of market demand or supply, and simultaneously placed smaller “genuine” orders to profit from price movement.
Profits: about $1.4 million over several months.
Legal Issues:
Charged under the anti-spoofing provisions of the Commodity Exchange Act.
The main question: Did Coscia intend to manipulate the market via the algorithm, or was it accidental?
Outcome:
Convicted on all 12 counts (6 spoofing, 6 commodities fraud).
Sentenced to 3 years in federal prison.
Appeal affirmed the conviction, establishing that directing an algorithm to manipulate markets constitutes criminal intent, even if the human is not entering orders manually.
Significance:
Demonstrates criminal liability for intentional misuse of algorithms.
Establishes that human designers/controllers are responsible for algorithmic actions.
2. Navinder Singh Sarao – “Flash Crash” Manipulation (UK/US, 2015)
Facts:
Sarao used an automated trading program to engage in layering and spoofing of E-mini S&P 500 futures.
His trades allegedly contributed to the 2010 Flash Crash in U.S. markets.
Large-scale market impact: his trading patterns amplified price drops.
Legal Issues:
Charges: commodities fraud, market manipulation, spoofing.
Key challenge: attributing algorithmic activity to human intent and proving market manipulation across borders.
Outcome:
Arrested in the UK; extradition proceedings followed to the U.S.
The case highlighted cross-jurisdictional accountability and the use of algorithms to manipulate markets.
Significance:
Shows how even individual traders using autonomous algorithms can cause systemic market disruptions and be held accountable.
Highlights regulatory and legal challenges in attributing criminal intent to algorithm-driven trades.
3. NSE Algo-Software Misconduct (India, 2016)
Facts:
NSE awarded a software contract to a vendor who used exchange data to develop algorithmic trading software for select traders.
The software gave certain traders an unfair advantage, violating market fairness.
This was not a “fat-finger” error but intentional misuse of exchange infrastructure.
Legal Issues:
Regulatory violations: unfair access and market manipulation.
Accountability extended to exchange officials and the vendor.
Question: Does misusing algorithmic software constitute criminal conduct, or is it regulatory/civil?
Outcome:
SEBI fined NSE and associated parties a total of ₹11 crore.
Criminal charges were considered in some instances, but primarily it was a regulatory enforcement case.
Significance:
Illustrates accountability for design and deployment of trading algorithms that advantage some market participants.
Shows that liability may extend beyond the trader to developers and exchange managers.
4. Citigroup Algorithmic Trading Error (UK, 2022)
Facts:
A trader entered a huge erroneous basket of equities (~$444 billion instead of ~$58 million).
The bank’s algorithm partially executed the erroneous trades, leading to ~$1.4 billion of equities sold before cancellation.
Market disruption occurred, though briefly.
Legal Issues:
Pure operational error rather than intentional misconduct.
Main question: Could negligence in supervising autonomous trading systems constitute criminal or regulatory liability?
Outcome:
FCA fined Citigroup £27.8 million; PRA fined £33.9 million.
No criminal prosecution for individuals, but severe regulatory sanctions enforced.
Significance:
Shows autonomous trading errors can cause huge financial and regulatory consequences even without criminal intent.
Highlights the importance of robust internal controls for algorithmic systems.
5. Knight Capital Trading Glitch (US, 2012)
Facts:
Knight Capital deployed a new algorithm that malfunctioned due to a configuration error.
Resulted in rapid buying and selling of ~4 million shares in 45 minutes, causing a $440 million loss.
The error almost bankrupted the firm.
Legal Issues:
Operational failure: Was this negligence or just an unfortunate error?
Could executives be held criminally liable for insufficient oversight of automated systems?
Outcome:
Regulatory and financial consequences were severe; Knight Capital had to secure emergency funding.
No known criminal convictions; primarily a case of operational failure and civil/regulatory accountability.
Significance:
A classic example of algorithmic error rather than intentional fraud.
Emphasizes that even unintentional autonomous trading errors have enormous consequences and require stringent controls.
6. Barings Bank Collapse – Nick Leeson (UK, 1995)
Facts:
Nick Leeson used derivatives trading systems to hide losses. While not a high-frequency algorithm, the trading platform allowed him to enter trades that caused massive exposure without oversight.
He eventually accumulated $1.3 billion in losses, bankrupting Barings Bank.
Legal Issues:
Criminal fraud: Did Leeson intentionally mislead his employer and regulators?
Automated systems enabled large trades with insufficient checks, raising the question of institutional accountability.
Outcome:
Leeson was convicted of fraud and sentenced to 6.5 years in prison.
Collapse highlighted the risks of unsupervised or poorly supervised trading systems, including early “algorithmic” exposure.
Significance:
Historical precedent showing that human intent combined with automated trading tools can cause catastrophic financial failures.
Provides context for accountability in modern autonomous trading errors.
Summary of Key Lessons from These Cases
Intentional Misconduct (Coscia, Sarao): Human designers/operators are criminally liable for algorithms used to manipulate markets.
Operational Errors (Citigroup, Knight Capital): Errors in autonomous trading cause regulatory liability; criminal liability requires recklessness or gross negligence.
Software Misuse / Design Faults (NSE, Barings): Developers, exchange officials, and traders may all be accountable if systems are used unfairly or without proper oversight.
Autonomy vs. Control: As algorithms become more autonomous, human liability increasingly depends on foresight, oversight, and compliance controls.
RELATED Blog
comments