The closest supervision model in existence today for Artificial Intelligence is that of Finance. This is due to many essential similarities.

Financial system

AI system

Common governance problem

Banks, brokers, exchanges

AI developers, model companies, deployers

Highly specialized private-sector institutions

Financial regulators

AI regulators/supervisors

Government needs oversight without knowing every technical detail

FINRA / exchanges / SROs

Potential AI SRO

Industry expertise can be incorporated into supervision

Capital requirements

AI safety/security requirements

Establish minimum risk-management requirements

Stress testing

AI safety evaluations / red-team testing

Test systems before they create unacceptable risks

Bank examinations

AI audits / technical assessments

Continuous examination rather than occasional legislation

Risk management frameworks

AI risk-management frameworks

Firms manage risks internally according to regulatory expectations

Regulatory reporting

AI incident / capability reporting

Government needs information that companies possess

Systemic financial risk

Systemic AI risk

Failure of one major institution can affect the wider system

Too-big-to-fail institutions

Frontier AI companies/models

A small number of actors can have disproportionate consequences

Market integrity

AI reliability/integrity

Public confidence is essential to system functioning

Consumer protection

AI consumer/user protection

Private incentives don't necessarily align with public interests

Insider trading / market manipulation

AI-enabled manipulation

Specialized rules needed for technologically sophisticated misconduct

Bank secrecy + regulatory access

Corporate IP + regulatory access

Supervisors need access to sensitive information without making it public

Federal financial regulators

Federal AI oversight

Government establishes the ultimate authority

SRO enforcement

Industry compliance/enforcement

Day-to-day enforcement can occur closer to the industry

Government backstop

Government AI backstop

State retains ultimate enforcement authority


Comments

Post a Comment

Popular posts from this blog

Image
Under the Joint Artificial Intelligence Governance model, frontier AI companies must demonstrate their ability to self-regulate. This is a recent example of actions towards achieving this goal. OpenAI, Anthropic probe tens of thousands of AI model security incidents As for its part of establishing standards, licensing, auditing and liability, government is in the stage of publicly probing AI companies nationally and internationally. These are sample probes into AI.  Senior leaders from Anthropic, OpenAI, Google and Meta testified under oath during a New York City Council hearing. U.S. FTC to Compel Testimony from AI Executives in Sweeping Probe of OpenAI and Anthropic U.S. Will Now Examine National Security Implications of New AI Models, Pre-Release OpenAI Faces Congressional Probe Over Swarm Hacking Incident Australia Summons OpenAI, Anthropic CEOs to AI Probe
Read more
Image
Financial Sector Governance could be the closest model to follow for Artificial Intelligence governance. How does the Financial sector work and how does it help in the case of AI? The financial-sector model essentially says: Private power + public authority + shared infrastructure + systemic risk requires joint governance . AI increasingly has exactly the same characteristics. Consider the parallel: Financial Sector AI Ecosystem Banks AI Companies Financial Markets AI Platforms/Models Capital Compute/Data/Investment Regulators AI Regulatory Authorities Central Banks Strategic/National AI Institutions Financial Infrastructure Cloud, Chips, Data Centers Systemic Financial Risk Systemic AI Risk Regulatory Capture AI Regulatory Capture Financial Stability AI/Strategic Stability Public/Private Governance Government–Industry AI Governance The most important lesson from Financial Sector Governance ...
Read more
Image
Artificial intelligence is developing along two major—and increasingly interconnected—paths: Commercial/Economic AI and Strategic AI. They require different governance philosophies. Commercial/Economic AI The guiding principle is: Regulate the risks while preserving competition and innovation. Its institutional foundations resemble those developed for other powerful economic systems: Financial regulation Antitrust law Consumer protection Environmental regulation Occupational safety Privacy law The objective is to allow innovation while preventing companies, markets, or technologies from imposing unacceptable costs on society. Strategic AI Strategic AI presents a fundamentally different challenge. Its guiding principle is: Maintain human control and strategic stability while preventing catastrophic misuse or escalation. Its institutional ancestors are therefore different: Nuclear arms control International humanitarian law Military command structures Intelligence oversight ...
Read more