Civil Law And AI-Controlled Financial Market Systemic Failure Claims In Europe

1. Introduction

AI is increasingly used in European financial markets for:

algorithmic and high-frequency trading;

portfolio management;

market-making;

fraud and market-abuse detection;

credit and liquidity assessment;

risk management;

automated order routing;

pricing;

collateral management;

stress testing;

investment recommendations;

financial forecasting.

This creates a distinctive civil-liability problem:

What happens when an AI-controlled financial system makes an error that is not confined to one investor but propagates through interconnected banks, trading venues, funds, clearing systems or other market infrastructure and contributes to a systemic market failure?

A conventional financial-loss claim normally concerns a bilateral relationship:

Bank → Investor → Loss

An AI-related systemic failure may instead look like:

AI model error → abnormal trading → market disruption → liquidity shock → clearing/settlement problems → multiple institutional losses → investor losses.

European law does not currently contain one single “AI systemic financial failure liability” cause of action. Instead, claims must generally be constructed from several areas of law:

MiFID II/MiFIR;

Market Abuse Regulation;

EMIR;

DORA;

EU financial-services legislation;

contract law;

tort/delict law;

product/service liability;

data-protection law where relevant;

corporate and directors' liability;

insolvency law;

national public-authority liability.

DORA is particularly important because it expressly recognises that financial-sector ICT interdependence can create systemic vulnerabilities and requires financial entities to maintain digital operational resilience. (EUR-Lex)

2. Meaning of an AI-Controlled Financial Market System

An AI-controlled financial market system may contain several connected layers:

The legal problem becomes much more complicated when these systems interact.

For example:

AI A increases buying.

→ AI B interprets the increased price as positive market information.

→ AI C increases leverage.

→ AI D reduces liquidity because volatility rises.

→ automated margin calls begin.

→ forced sales occur.

→ prices collapse.

The resulting loss may be described as a systemic AI-induced market event.

3. What Is a Systemic Failure?

A systemic failure is broader than an ordinary software error.

Ordinary AI error

One investment recommendation is wrong.

Systemic AI failure

The AI error affects interconnected market participants and produces cascading consequences.

Potential examples include:

simultaneous algorithmic selling;

incorrect volatility calculations;

erroneous liquidity requirements;

automated margin spirals;

correlated portfolio decisions;

faulty risk models;

defective market data;

AI manipulation of prices;

erroneous execution across multiple venues;

widespread model instability;

common dependence on one AI vendor;

failure of a critical cloud/ICT provider.

DORA specifically recognises the significance of ICT interdependencies and the possibility that disruption involving critical ICT providers can threaten the stability and integrity of the Union financial system. (EUR-Lex)

4. European Regulatory Framework

A. MiFID II

MiFID II regulates investment services and financial markets.

For AI-controlled investment systems, important principles include:

investor protection;

suitability;

appropriateness;

conduct of business;

organisational requirements;

risk management;

best execution;

record keeping;

conflicts of interest.

AI does not remove the regulated firm's obligations.

If an investment firm delegates a decision to an algorithm, the firm generally cannot simply argue:

“The algorithm made the decision, therefore nobody is responsible.”

5. DORA — Digital Operational Resilience Act

Regulation (EU) 2022/2554 is particularly relevant.

DORA establishes requirements concerning:

ICT risk management;

ICT incident reporting;

resilience testing;

ICT third-party risk;

operational continuity;

cyber risks;

critical ICT providers.

It requires financial entities to maintain a comprehensive ICT-risk framework and identify interdependencies between their systems and ICT providers. (EUR-Lex)

DORA also requires systems to be sufficiently reliable and capable of processing necessary data and handling peak transaction volumes, including under stressed market conditions. (EUR-Lex)

Importance for AI

If an AI trading system fails because:

it cannot handle extreme market volume;

its data pipeline becomes corrupted;

its model behaves unpredictably;

a critical ICT supplier fails;

inadequate testing was performed;

DORA may become highly relevant to the regulatory and civil-liability analysis.

However:

A DORA breach does not automatically equal a private damages award.

The claimant must still identify the applicable private-law remedy, loss and causation.

6. Market Abuse Regulation

The Market Abuse Regulation is relevant where AI is involved in:

insider dealing;

unlawful disclosure;

market manipulation;

misleading transactions;

manipulation of benchmarks or prices.

An AI system could potentially execute thousands of transactions in a very short period.

The legal question is not merely:

“Was the trader human?”

Instead, authorities may investigate:

who designed the algorithm;

who deployed it;

who controlled its parameters;

what information was available;

whether warning signals existed;

whether controls were adequate;

whether the conduct produced artificial prices.

7. EMIR and Clearing-System Risk

Where AI-controlled activity affects derivatives, clearing and margining, EMIR becomes important.

Potential systemic pathways include:

AI trading error

→ derivative-price movement

→ margin requirement increases

→ forced liquidation

→ additional price movement

→ further margin calls.

This can create a feedback loop.

Consequently, liability analysis may involve:

investment firms;

counterparties;

central counterparties;

clearing members;

trading venues;

data providers;

technology suppliers.

8. AI Act

The EU AI Act provides a general AI regulatory framework.

However, it is important not to assume that every AI trading system is automatically a high-risk AI system under the AI Act.

The AI Act's classification depends upon the system's specific intended purpose and applicable provisions.

Financial-sector regulation therefore remains essential.

The practical relationship can be expressed as:

AI Act + DORA + MiFID II/MiFIR + MAR + EMIR + national private law

rather than:

AI Act alone = complete financial AI liability regime.

9. Case Law

Because fully autonomous AI-generated systemic market failures are a relatively new phenomenon, European courts do not yet have a mature body of cases specifically deciding:

“AI trading system caused systemic financial collapse.”

Accordingly, the most useful authorities are financial-market cases whose principles can be applied to AI systems, together with cases concerning State liability and financial-market regulation.

The distinction between direct AI precedent and analogical authority is therefore important.

10. Case 1 — Spector Photo Group NV and Van Raemdonck v CBFA, C-45/08

Court: CJEU
Date: 23 December 2009
ECLI: EU:C:2009:806

The case concerned insider dealing and the interpretation of EU market-abuse rules. The CJEU examined the conditions under which transactions can fall within the insider-dealing prohibition. (Infocuria)

Importance for AI

Imagine an AI trading system receives information that constitutes inside information and automatically executes trades.

The fact that the transaction was generated by an algorithm does not make the market-abuse rules irrelevant.

The investigation may have to establish:

what information entered the system;

whether the information was inside information;

how the algorithm processed it;

whether the trading strategy was designed or modified because of that information;

who controlled the system.

Legal principle

Automated execution does not automatically eliminate responsibility under market-abuse rules.

11. Case 2 — Geltl v Daimler AG, C-19/11

Court: CJEU
Date: 28 June 2012
ECLI: EU:C:2012:397

The case concerned the meaning of inside information, including intermediate steps in a protracted process.

The Court explained the circumstances in which intermediate developments can constitute sufficiently precise information for market-abuse purposes. (Infocuria)

AI relevance

AI systems increasingly operate on:

partial information;

probability;

predicted events;

intermediate developments.

An AI system may identify a corporate event before it becomes publicly known.

The legal issue then becomes:

Does the information processed by the system satisfy the legal definition of inside information?

Systemic significance

If multiple AI systems simultaneously react to the same undisclosed information, the resulting automated trading can amplify the market impact.

Thus:

information → AI prediction → automated trades → price movement

can become a market-abuse issue as well as a civil-liability issue.

12. Case 3 — Genil 48 SL v Bankinter, C-604/11

Court: CJEU
Date: 30 May 2013
ECLI: EU:C:2013:344

The case concerned MiFID conduct-of-business requirements, including suitability and appropriateness assessments in connection with investment services and interest-rate swaps. (Infocuria)

Importance for AI

Suppose an AI system automatically recommends or executes complex derivatives for clients.

The financial institution cannot necessarily avoid suitability/appropriateness obligations merely because the recommendation was algorithmically generated.

Potential claim

A claimant could argue:

The firm delegated investment decisions to an AI system whose methodology was unsuitable for the client's circumstances and failed to satisfy applicable conduct-of-business obligations.

Systemic connection

If the same flawed AI model is deployed across thousands of clients, an individual suitability failure can become a mass-scale risk.

13. Case 4 — Petruchová v FIBO Group Holdings, C-208/18

Court: CJEU
Date: 3 October 2019
ECLI: EU:C:2019:825

The case concerned a retail investor trading on the international foreign-exchange market through a brokerage company.

The CJEU treated the investor as a consumer for the purposes of the relevant jurisdictional rules where the transactions were outside the person's professional activity. (Infocuria)

AI relevance

AI-controlled trading increasingly involves retail investors through:

automated platforms;

algorithmic portfolio tools;

CFDs;

automated execution;

robo-advisory systems.

The classification of the customer matters because it can affect:

jurisdiction;

consumer protection;

applicable procedural rules;

access to courts.

Principle

Sophisticated financial technology does not automatically transform a retail customer into a professional market participant.

14. Case 5 — Kolassa v Barclays Bank, C-375/13

Court: CJEU
Date: 28 January 2015
ECLI: EU:C:2015:37

The case concerned an investor's claim connected with an investment product issued by a bank and questions of jurisdiction.

The CJEU examined where a consumer could bring proceedings concerning investment losses. (Infocuria)

AI relevance

Systemic AI failures may involve numerous defendants:

investment bank;

trading platform;

AI vendor;

data provider;

fund;

broker.

Jurisdiction becomes particularly complicated when the system operates across several Member States.

Practical principle

A claimant should analyse:

place of domicile;

place of damage;

contractual jurisdiction;

consumer jurisdiction;

place where relevant financial services were provided.

15. Case 6 — Alpine Investments BV v Minister van Financiën, C-384/93

Court: CJEU
Date: 10 May 1995
ECLI: EU:C:1995:126

The case concerned restrictions on cold calling for financial services involving commodities futures.

The CJEU recognised investor protection and confidence in financial markets as legitimate public-interest considerations capable of justifying restrictions, subject to proportionality. (Infocuria)

AI relevance

This case supports an important proposition:

Financial-market regulation can legitimately impose restrictions on technological or commercial practices when necessary to protect investors and market integrity.

Applied to AI, this may support regulatory requirements concerning:

automated trading;

algorithmic marketing;

risk controls;

system testing;

AI deployment.

Systemic significance

Investor protection is not limited to individual contractual loss.

Market integrity itself is a legitimate regulatory objective.

16. Case 7 — Köbler v Republik Österreich, C-224/01

Court: CJEU
Date: 30 September 2003
ECLI: EU:C:2003:513

Köbler is not a financial-market case. It is important for the State-liability dimension.

The CJEU established that a Member State can, under the conditions of EU law, be liable for damage caused by sufficiently serious infringements of EU law attributable to a final national court. The Court identified rights conferred on individuals, a sufficiently serious breach and direct causal connection as central elements. (Infocuria)

AI-systemic relevance

Suppose:

EU financial rules impose mandatory supervisory obligations;

a national authority seriously fails to implement or apply them;

an AI-driven systemic event occurs;

investors suffer losses.

The claimant might investigate whether a State-liability claim is legally available.

But this is a demanding route.

A regulatory failure does not automatically create State liability.

17. Case 8 — Traghetti del Mediterraneo v Italy, C-173/03

Court: CJEU Grand Chamber
Date: 13 June 2006
ECLI: EU:C:2006:391

The Court further developed the principle of Member State liability for breaches of EU law.

The case concerned restrictions on national rules limiting State liability.

AI-financial relevance

It is useful when systemic failure allegations extend beyond private financial firms to:

regulators;

supervisory authorities;

national courts;

implementation failures.

The case confirms that national procedural rules cannot simply eliminate the effectiveness of EU-law State liability. (Infocuria)

18. Case 9 — Schindler Holding and Others / Financial Regulatory Jurisprudence

European financial-market jurisprudence also demonstrates that financial regulation frequently balances:

market integrity;

investor protection;

free movement;

proportionality;

regulatory supervision.

This becomes important for AI because regulators may legitimately impose restrictions on high-risk automated systems even where firms argue that such restrictions interfere with technological or commercial freedom.

19. Who Could Be Liable?

A systemic AI failure may involve several potentially responsible actors.

1. Financial institution

For:

negligent deployment;

inadequate testing;

inadequate monitoring;

inadequate governance;

failure to control algorithmic risk.

2. AI developer

Potential liability may arise depending upon:

contract;

applicable product/service liability rules;

negligent design;

defective software;

misleading documentation.

3. Data provider

Potentially relevant where:

market data were incorrect;

data were delayed;

corrupted data were supplied;

essential information was incomplete.

4. Cloud/ICT provider

Particularly important under DORA where critical ICT dependencies exist.

5. Trading venue

Potential liability may arise depending upon:

contractual terms;

regulatory obligations;

operational failures;

market infrastructure responsibilities.

6. Central counterparty

Potentially relevant where clearing or margin mechanisms contribute to the loss.

7. Directors and management

Possible liability may arise under national company law where management failed to establish appropriate risk controls.

8. Public regulator

Possible State-liability claims are legally distinct and subject to demanding conditions.

20. The Central Problem: Causation

Causation is likely to be the hardest issue.

Consider:

AI error

↓

abnormal orders

↓

price volatility

↓

liquidity withdrawal

↓

margin calls

↓

forced selling

↓

market crash

↓

investor losses

Which event legally caused the claimant's loss?

The defendant may argue:

“The market was already volatile.”

Another defendant may say:

“The claimant's loss resulted from macroeconomic conditions.”

Another may say:

“Our AI merely responded to market conditions.”

Therefore, causation must be reconstructed carefully.

21. Counterfactual Causation

A claimant may ask:

What would have happened if the AI system had operated correctly?

Possible counterfactuals include:

Scenario A

Without the AI error, no systemic event occurs.

Scenario B

The AI error occurs but another independent event would still have caused the crash.

Scenario C

The AI error accelerates an existing crisis.

Scenario D

The AI error increases the magnitude of the crisis but does not cause it.

Scenario E

Several AI systems independently contribute.

The legal outcome can differ substantially between these scenarios.

22. Multiple AI Systems and Concurrent Causation

Suppose five institutions use similar AI models.

All five systems respond to the same market signal.

The resulting transactions amplify the market movement.

Now:

Which institution caused the loss?

Possible legal approaches may involve:

concurrent causation;

joint causation;

contribution between defendants;

proportionate liability;

national rules concerning indivisible damage;

contractual allocation of risk.

The precise solution depends heavily upon national private law.

23. Foreseeability

Foreseeability is another important issue.

A financial institution may argue:

“Nobody could have predicted this AI behaviour.”

The claimant may respond:

“The failure was foreseeable because the system was known to be highly correlated with other automated systems.”

DORA strengthens the importance of anticipating ICT risks because it expressly requires financial entities to identify ICT risks, dependencies and critical functions. (EUR-Lex)

Therefore, evidence of prior warnings, stress tests, failed simulations or known model weaknesses may become important.

24. AI Model Risk

Financial AI creates a distinctive category of model risk.

Examples:

training-data bias;

overfitting;

regime-change failure;

incorrect assumptions;

feedback loops;

hallucinated financial information;

unstable optimisation;

excessive correlation;

insufficient stress testing;

unexpected interaction between algorithms.

The legal question becomes:

Was the model risk reasonably foreseeable and properly managed?

25. Feedback Loops

Feedback loops are especially important.

For example:

AI predicts price decline

→ AI sells

→ price falls

→ market data confirms AI's prediction

→ AI sells more

→ price falls further.

This creates a self-reinforcing cycle.

If multiple market participants use similar models, the loop can become systemic.

The relevant liability question is therefore not only:

“Was the prediction wrong?”

but also:

“Was the deployment architecture reasonably designed to prevent destabilising feedback?”

26. Common-Model Risk

Systemic risk increases when many financial institutions rely upon:

the same AI provider;

the same model;

the same data source;

the same cloud infrastructure;

the same market signals.

This is a form of concentration risk.

DORA specifically addresses ICT third-party risk and requires financial entities to assess dependencies on ICT providers supporting critical or important functions. (EUR-Lex)

27. Contractual Liability

Contracts may exist between:

bank and AI provider;

broker and trading platform;

investor and broker;

fund and technology supplier;

exchange and technology vendor.

Important contractual clauses include:

warranties;

service levels;

uptime;

accuracy;

liability caps;

exclusions;

indemnities;

force majeure;

audit rights;

incident reporting;

cybersecurity obligations.

A liability cap may become controversial where:

the supplier committed gross negligence;

mandatory law prevents exclusion;

the clause violates applicable regulatory requirements;

the loss is outside the contractual risk allocation.

The enforceability of such clauses depends on applicable national law.

28. Tort/Delict Liability

A claimant may potentially establish:

duty of care;

breach;

damage;

causation;

foreseeability, where required;

absence of applicable defence.

The duty may involve:

reasonable algorithm testing;

risk monitoring;

cybersecurity;

human oversight;

emergency shutdown;

model validation;

market-abuse controls.

29. Regulatory Breach Does Not Automatically Equal Civil Damages

This distinction is essential.

For example:

DORA breach → investor automatically receives damages.

That conclusion does not necessarily follow.

The claimant must examine:

whether the provision creates an individual right;

whether national law provides a private cause of action;

whether the claimant suffered legally recognised damage;

whether the breach caused that damage;

whether other causes intervened.

The CJEU's State-liability jurisprudence similarly requires specific conditions rather than treating every EU-law violation as an automatic damages claim. Köbler is an important illustration. (Infocuria)

30. Regulatory Failure Claims

A particularly difficult situation arises when the claimant argues:

“The regulator should have stopped the AI system earlier.”

This requires distinguishing:

Private firm's negligence

The institution deployed an unsafe AI model.

from:

Supervisory failure

The regulator allegedly failed to exercise its statutory powers appropriately.

from:

Legislative failure

The legal framework allegedly did not adequately address emerging AI risks.

These are legally different claims.

31. Evidence in AI-Systemic Failure Litigation

The following evidence can become critical:

32. Expert Evidence

AI systemic litigation will often require experts in:

financial economics;

quantitative finance;

algorithmic trading;

software engineering;

AI/model validation;

cybersecurity;

market microstructure;

risk management;

accounting;

causation.

An expert may need to reconstruct:

Input → Model → Output → Trading behaviour → Market reaction → Loss

This is much more complicated than ordinary contractual causation.

33. Damages

Potential damages may include:

Direct investment loss

Loss caused by an erroneous transaction.

Lost profits

Profits that would allegedly have been obtained absent the AI failure.

Transaction costs

Costs resulting from unnecessary trading.

Liquidity losses

Loss resulting from forced liquidation.

Hedging losses

Losses caused by defective risk calculations.

Business interruption

For financial institutions unable to operate.

Reputational damage

Potentially recoverable depending on national law.

Systemic market losses

Much more difficult because the claimant must distinguish its legally recoverable loss from general market decline.

34. Limitation and Allocation of Risk

A financial AI contract may contain:

maximum liability;

exclusion of consequential loss;

exclusion of lost profits;

force-majeure clauses;

market-volatility exclusions.

Courts must determine whether such clauses apply and whether mandatory statutory rules restrict them.

The fact that the financial loss is enormous does not itself invalidate a contractual limitation.

35. Force Majeure and Extraordinary Market Events

Defendants may argue:

“The market crash was an extraordinary event beyond reasonable control.”

But an AI failure may be harder to characterise as force majeure if:

the system was inadequately tested;

the failure was foreseeable;

warnings existed;

risk controls were inadequate;

the institution failed to activate emergency controls.

The distinction between external market shock and internally generated technological failure is therefore important.

36. Systemic AI Failure and DORA

DORA's framework is particularly relevant because it requires:

ICT risk management;

continuous monitoring;

resilience;

incident management;

testing;

third-party risk management;

identification of critical dependencies.

The regulation also requires incident processes to identify, document and address root causes so as to prevent recurrence. (EUR-Lex)

Thus, in litigation, a claimant may ask:

Was the AI system governed as a critical operational risk rather than merely as an investment tool?

37. AI Failure vs Market Risk

This distinction is fundamental.

Ordinary market risk

Investor loses money because:

stock prices fall;

interest rates change;

currency depreciates;

market conditions deteriorate.

AI-induced operational risk

Investor loses money because:

AI incorrectly trades;

data are corrupted;

algorithm malfunctions;

AI causes abnormal order flows;

risk controls fail.

AI-amplified market risk

A market movement begins normally but AI systems amplify it.

The third category may be the most legally difficult.

38. Practical Liability Matrix

39. Important Case-Law Principles

40. A Model Hypothetical

Assume a European investment bank deploys an AI trading model.

The model incorrectly interprets a market-data feed.

It automatically:

sells €10 billion of securities;

triggers price declines;

causes other AI systems to sell;

increases volatility;

triggers margin calls;

causes forced liquidation;

affects a central counterparty;

creates liquidity shortages;

causes losses for thousands of investors.

Potential claims

Against the bank

inadequate model validation;

inadequate stress testing;

inadequate risk limits;

DORA compliance failures;

negligence;

contractual breach.

Against the AI supplier

defective software;

contractual breach;

failure to warn;

negligent design.

Against the data supplier

inaccurate market data;

contractual breach;

negligent data provision.

Against other market participants

Possible contribution claims depending on national law.

Against the regulator

Potential State-liability analysis if statutory EU obligations were seriously breached and the demanding conditions for State liability are satisfied.

41. Causation in the Hypothetical

The claimant must distinguish:

AI error

from

market volatility

from

independent macroeconomic factors

from

actions of other algorithms

from

clearing-system responses.

A sophisticated expert model may therefore be required.

The question is not simply:

“Did the AI make a mistake?”

It is:

“Did the legally attributable AI-related breach materially cause the claimant's recoverable loss?”

42. Defences

Potential defendants may rely upon:

1. Market volatility

The loss was caused by ordinary market conditions.

2. Intervening cause

Another institution or market event caused the loss.

3. Contributory negligence

The claimant accepted excessive risk.

4. Contractual limitation

The agreement limits liability.

5. Force majeure

The event was extraordinary and uncontrollable.

6. Regulatory compliance

The institution complied with applicable rules.

7. Lack of causation

The AI error did not materially cause the claimed loss.

8. Remoteness

The claimed systemic loss was too remote.

43. Special Problem of AI Explainability

AI systems can be difficult to explain because the model may involve:

neural networks;

reinforcement learning;

adaptive parameters;

multiple data feeds;

continuously changing market conditions.

But legal responsibility cannot simply disappear because the technology is complex.

For financial institutions, the critical question becomes:

Can the institution demonstrate what the AI was designed to do, what controls existed, what risks were known and what happened during the incident?

DORA's emphasis on documented ICT-risk frameworks, incident management and governance makes this particularly important. (EUR-Lex)

44. Civil-Law Liability Formula

A useful formula is:

AI System + Financial-Service Duty + Defective/Unreasonable Deployment + Regulatory/Contractual/Tort Breach + Causation + Recoverable Damage = Potential Civil Liability

For systemic claims:

AI Error + Interconnected Market Systems + Foreseeable Systemic Risk + Inadequate Controls + Causal Contribution to Market Disruption + Individual Loss = Potential Systemic AI Liability Claim

For regulatory claims:

EU-Law Obligation + Serious Breach + Individual Damage + Direct Causal Link = Potential State-Liability Claim

The last formula reflects the general EU State-liability framework illustrated by Köbler. (Infocuria)

45. Key Legal Principles

AI does not become a legal person merely because it makes autonomous financial decisions.

The financial institution normally remains legally responsible for the regulated activity it conducts through technology.

DORA makes digital operational resilience a central part of financial-sector risk management. (EUR-Lex)

Systemic AI failure requires analysis of interconnected causation, not merely individual software error.

Market-abuse rules can apply to algorithmically executed transactions.

AI investment advice remains subject to applicable investor-protection obligations.

Retail investors may retain consumer protections even when using sophisticated financial platforms.

A regulatory violation does not automatically establish a private damages claim.

State liability is legally distinct from private financial-firm liability.

Expert evidence is likely to be crucial in proving AI causation.

Common AI models and common ICT providers can create concentration and systemic-risk problems.

Contractual liability limitations must be assessed under applicable national law and mandatory financial legislation.

46. Exam-Oriented Conclusion

AI-controlled financial markets create a new form of civil liability in which technological error, financial regulation and systemic causation overlap.

The most important legal question is not simply whether an AI system malfunctioned. The court must determine:

Who controlled the AI, what legal duty applied, whether the system was reasonably designed and monitored, whether the risk was foreseeable, whether regulatory requirements were breached, whether the AI event materially contributed to the systemic disruption, and what portion of the claimant's loss is legally recoverable.

The strongest financial-market authorities include Spector Photo Group, Geltl, Genil 48, Petruchová, Kolassa and Alpine Investments, while Köbler and Traghetti del Mediterraneo provide important principles for claims involving public-authority or supervisory failures. These cases are mostly analogical rather than direct AI-systemic-failure precedents, because European courts have not yet developed a substantial body of judgments concerning an autonomous AI system causing a systemic financial-market collapse. (Infocuria)

Ultra-short revision keywords

AI trading – algorithmic trading – systemic risk – market failure – MiFID II – MiFIR – MAR – EMIR – DORA – ICT risk – operational resilience – model risk – algorithmic feedback loop – liquidity spiral – market manipulation – insider information – suitability – investor protection – causation – concurrent causation – foreseeability – contractual liability – tort liability – vendor liability – third-party ICT risk – State liability – damages – expert evidence – systemic financial loss.