Operational Risk Dominates Adversarial Risk In Market Infrastructure
Sources: 1 • Confidence: Medium • Updated: 2026-04-11 19:04
Key takeaways
- Blankfein said the technological risk that worries him most is unintentional failure (e.g., human error) rather than primarily malevolent state actors.
- Blankfein said he expects AI to automate much of what banks and other white-collar workers do, with many roles changing or disappearing while some hands-on service work remains comparatively insulated for longer.
- Blankfein said that in a major downturn governments would still need to stabilize banks because the banking system is the main transmission channel for monetary and fiscal stimulus to the public.
- Blankfein said Goldman’s risk management relied on independent marking of positions and forcing risk-takers to validate higher valuations by selling assets at those prices.
- Blankfein said COVID highlighted that domestic control of strategic manufacturing (e.g., vaccines and PPE) influences access in a crisis, reinforcing pressures toward localized supply chains.
Sections
Operational Risk Dominates Adversarial Risk In Market Infrastructure
- Blankfein said the technological risk that worries him most is unintentional failure (e.g., human error) rather than primarily malevolent state actors.
- Blankfein argued that adding many layers of checks can backfire because responsibility diffuses and routine checking becomes mind-numbing, so people stop taking any single control seriously.
- Blankfein said the shift from open out-loud trading floors to digital communication and algorithmic trading reduces collective error-detection because fewer people can sanity-check questionable actions in real time.
- Blankfein framed technology as leverage that amplifies outcomes, improving performance when aligned and worsening damage when misapplied or failing.
- Blankfein recounted a software-testing error that briefly triggered selling all stocks starting with L through P for one dollar, generating roughly $1.5–$2.0 billion of transactions in about 15 seconds that were mostly unwound.
- Blankfein argued that as technology progresses, accidents can have vastly larger potential casualty and damage footprints than earlier industrial accidents.
Banking Is Already An Engineering/Latency Business; Ai Shifts Labor But Not Accountability
- Blankfein said he expects AI to automate much of what banks and other white-collar workers do, with many roles changing or disappearing while some hands-on service work remains comparatively insulated for longer.
- Blankfein predicted AI-driven productivity could lead society toward progressively shorter workweeks over time, regardless of whether people prefer that transition.
- Blankfein said many time-intensive tasks in finance and law have already been largely eliminated by technology.
- Blankfein argued AI can analyze data and simulate probabilities but cannot truly 'take risk,' so human judgment remains necessary for key financial decisions.
- Blankfein said engineers made up over a third of Goldman’s workforce during his tenure and that modern market-making is a millisecond competition where even physical proximity of computers can determine outcomes.
Crisis Plumbing: Why Governments Backstop Banks
- Blankfein said that in a major downturn governments would still need to stabilize banks because the banking system is the main transmission channel for monetary and fiscal stimulus to the public.
- Blankfein described a credit crisis as a daisy-chain payment and solvency uncertainty problem that can freeze the system until a large balance sheet (typically government) temporarily guarantees payments to restart flows.
- Blankfein estimated that during the 2008 crisis there was roughly a 15% to 20% chance of a longer-lasting breakdown.
- Blankfein said he expects that if a crisis like 2008 recurred, the government would act quickly despite political polarization to unfreeze the system.
Risk Management As Price Discovery + Early Hedging + Incentive Alignment
- Blankfein said Goldman’s risk management relied on independent marking of positions and forcing risk-takers to validate higher valuations by selling assets at those prices.
- Blankfein said that in emerging stress Goldman would limit directional exposure by pausing risk accumulation until it could source the other side of trades, and would buy market insurance early while it was cheap when it could not.
- Blankfein argued Goldman’s partnership-like culture—pay tied to firm-wide outcomes and broad owner oversight—helps surface bad behavior and manage risk even if it slows decision-making.
Deglobalization Drivers: Regulatory Ring-Fencing + Strategic Supply Resilience
- Blankfein said COVID highlighted that domestic control of strategic manufacturing (e.g., vaccines and PPE) influences access in a crisis, reinforcing pressures toward localized supply chains.
- Blankfein said the global financial crisis pushed regulators and governments to focus on the jurisdictional location of assets and liabilities, moving finance toward more national financial thinking.
- Blankfein argued that globalization and deglobalization are cyclical rather than a one-time regime shift.
Watchlist
- Blankfein warned that private credit and other private assets are hard to price due to illiquidity and can become problematic if expected returns do not adequately compensate for that illiquidity.
- Blankfein said the technological risk that worries him most is unintentional failure (e.g., human error) rather than primarily malevolent state actors.
Unknowns
- What is the time horizon and implementation detail behind the reported 100% equity allocation (e.g., strategic long-term allocation vs tactical positioning; any hedges)?
- What specific metrics or observations would substantiate the claim that private-credit returns may not compensate for illiquidity (e.g., observed secondary discounts, NAV adjustments, gating behavior)?
- How large is the actual exposure of retail channels (401(k)-like vehicles) and insurance-company balance sheets to private credit in the contexts being discussed here?
- What concrete governance/automation replaces the lost 'trading floor' collective error-detection mechanism in electronified markets, and how effective is it?
- What are the documented details of the trade-error anecdote (venue, system, reversal rules), and how representative is it of broader operational risk frequency?