Quis custodiet ipsos custodes – ‘Who watches the watchers?’ A question as pertinent to bank regulation today as when Plato raised the topic in his Republic in 370BC, or when Juvenal subsequently penned the above quote in his Satires in the second century AD.
This is particularly relevant given recent comments from Michelle Bowman, Vice Chair for Supervision at the Federal Reserve (Fed), on the failure of Silicon Valley Bank (SVB), alongside the initial findings of the Starling Advisory Group’s independent review1.
What happened with SVB
SVB’s vulnerabilities predated its failure in March 2023. Between 2019 and 2021, the bank’s assets almost tripled, from $71bn to $211bn, with much of the deposit growth invested in longer-duration securities. As interest rates rose, the value of these assets fell, creating unrealised losses. By the end of 2022, these losses exceeded the bank’s capital base, while 94% of deposits were uninsured and concentrated among specific clients.
Following the announcement of a $1.8bn loss on securities sales and plans to raise $2.25bn of capital, depositors withdrew more than $40bn (around 25% of the deposit base) on 9 March. With a further $100bn of withdrawals expected the following day, regulators closed the bank on 10 March2.
| Unrealised securities losses as a share of adjusted equity3 |
![]() |
| Source: Probative Inquiry 2023 Financial Sector Stress Events, Starling Advisory Group; September 2026 |
What was the Fed’s role?
Crucially, the Fed was aware of many of these risks.
From 2019 onwards, the Fed issued 54 supervisory findings against SVB. When the bank failed in March 2023, 31 remained open. Three times the number at peer firms. Deficiencies in SVB’s interest rate risk management were identified in 2020, 2021 and 2022.
However, these findings did not result in sufficiently timely action. Fed supervisors, for example, took more than seven months to develop an informal enforcement action covering governance, risk management and liquidity; the bank failed before it could be issued.
Bowman’s recent comments offer a possible explanation for the delay: “a long-standing culture of risk aversion” 4. She also noted that “…staff believed it was personally safer to take no action unless they were certain the action was exactly right”.
Casting our minds back to Plato’s Republic, Socrates compared the ideal guardians to well-bred watchdogs: gentle towards those they protected, but fierce when confronted by a threat. To us, the Fed’s supervisory culture appears to have discouraged the latter.
Improvements to oversight
Changes, however, have since been introduced. Supervisors are now directed to focus more closely on material financial risks, and less on procedural or documentation shortcomings, while the range of tools available to them has been expanded. They must also report monthly to senior supervisory officials where they are uncertain whether action is warranted, allowing concerns to be escalated rather than delayed. In Plato's terms, the intention is to restore some of the watchdog's bite.
The impact for financial credit
We believe that for bank creditors, these changes, with earlier supervisory intervention, should be supportive. Our equity colleagues may grumble at requirements to raise capital, strengthen liquidity, adjust underwriting standards or restrict distributions, but creditors typically benefit when vulnerabilities are addressed early, particularly while banks retain a wider range of options. The situation with SVB illustrates this point. Its vulnerabilities stemmed not from deteriorating asset quality but from interest rate and liquidity risk. Earlier intervention to reduce duration risk, bolster liquidity or raise capital may have come at a cost to profitability and shareholders, but could also have left the bank substantially more resilient.
[1] Probative Inquiry 2023 Financial Sector Stress Events, Starling Advisory Group, 21 September 2026
[3] Equity capital plus tax-adjusted available for sale (AFS) unrealised losses; mean over banks with $10bn in assets or more as of Q4 2022













