Bank of America shares sank on Wednesday after the lender reported third-quarter revenue below expectations and absorbed a $1.2 billion penalty from U.S. regulators. In the same round of commentary, CEO Brian Moynihan disclosed that the bank had put 1,000 outside contractors to work patching software weaknesses exposed by Anthropic’s Mythos, a rare public accounting of what defending against advanced AI can cost a major bank.
The stock fell about 8 percent in early trading, according to CNBC, and traded nearly 10 percent lower at one point before recovering part of the loss. It closed at $51.39, down 4.76 percent, on volume more than three times its three-month average, based on market data reported by The Motley Fool. That intraday slide briefly marked the bank’s biggest one-day drop since April 2025.
The revenue shortfall came mostly from Wall Street trading. Bank of America reported about $30 billion in revenue, roughly $570 million below analyst expectations, with trading misses in fixed income and equities totaling about $750 million. Fixed income revenue fell 17 percent to $2.55 billion on weakness in mortgages, municipal bonds and credit, around $450 million under estimates. Equities trading rose 10 percent to $2.51 billion, but that was still nearly $300 million short of forecasts. Investment banking revenue declined 11 percent to $1.8 billion, close to what management had signalled in September. The bright spot was net interest income, a closely watched profit driver, which rose about 8 percent and edged past estimates.
The regulatory penalty drove most of the earnings miss. A day before the results, the Treasury Department announced a $1.2 billion civil penalty for failures in anti-money-laundering and sanctions controls. Reporting on the action said it was imposed by the Office of the Comptroller of the Currency and the Financial Crimes Enforcement Network, which cited weaknesses in the bank’s Bank Secrecy Act safeguards. Treasury asserted that the bank had effectively outsourced significant compliance functions. The charge is not tax-deductible and reduced earnings by 17 cents per share. Excluding it, adjusted earnings came to $1.08 per share, two cents under the $1.10 consensus, which ended a streak of six consecutive quarterly beats.
Analysts pressed Moynihan on the call about earlier guidance that compliance-related costs would be contained, and several remarked on the sharp fall in the share price. Later, speaking to CNBC, Moynihan did not commit to a timeline for resolving the bank’s regulatory issues. One analysis noted that a second enforcement action raises the risk of heavier compliance spending going forward.
The most unusual disclosure of the day had nothing to do with trading desks. Moynihan told CNBC that Bank of America brought in 1,000 contractors during the second quarter to help patch software flaws, and that they are still there, working alongside roughly 3,000 employees already assigned to the task. He used the phrase “Mythos remediation” on the earnings call, and he said the vulnerabilities spanned both the bank’s own code and software supplied by vendors. “That risk is real risk,” he said. Because the 1,000 workers were hired as consultants, he noted, they are not counted in the bank’s headcount.
Mythos is an Anthropic model that can find and exploit weaknesses in software. The company released it in April to a small group of banks, technology firms and government agencies through a programme called Project Glasswing, with the aim of letting defenders fix flaws before attackers can use them. Bank of America is among the lenders with access. CNBC said Moynihan’s remarks are among the first by a major U.S. corporation to put numbers on the staffing and spending needed to protect systems against such a model. JPMorgan’s Jamie Dimon has also reported hundreds of employees working full-time on the problem, and regulators in the United States and Europe have been paying closer attention to AI-related cyber risk at banks.
The price tag matters, but it needs careful reading. Moynihan said the bank spent a “couple hundred million plus” in the quarter, and that figure covers two separate efforts, the software remediation and repairs to anti-money-laundering controls. It cannot be treated as the standalone cost of dealing with Mythos. He expects the Mythos share to decline over time, reasoning that once systems are patched there is less left to patch. The bank also argues that AI is paying for itself elsewhere. CNBC reported that its roughly 130 internal AI projects are expected to deliver about $700 million in savings next year, so the technology is both a cost and a source of efficiency.
Management also offered some reassurance on the underlying business. The bank guided for net interest income to grow about 8 percent in 2027, up from an earlier forecast of 5 to 7 percent. Some commentary on the day argued that, apart from the fine, it was a respectable quarter and that the stock’s reaction looked specific to Bank of America, since JPMorgan Chase and Wells Fargo held roughly steady. Others saw the combination of a trading shortfall and a large enforcement action as a reason for caution, because regulatory problems can take years to work through and can restrict growth in the meantime.
For investors, the open questions are practical. How long will compliance and remediation costs stay elevated, and will regulators attach further conditions to the bank’s operations? Will trading revenue recover, particularly in fixed income, which has been the weaker half of the desk? And will the spending on software patching fall as quickly as management hopes? The answers will show up over the next few quarters in expense lines rather than in any single announcement.
The episode also shows how artificial intelligence is changing the economics of banking from two directions at once. It can trim costs and speed up routine work, yet the same class of tools can uncover decades-old weaknesses and force an expensive scramble to close them. Readers who follow how AI is reshaping finance and cybersecurity can find related coverage at Innovation Times.
