How have banks responded to AI risk when nothing went wrong?
Not every AI story at a bank is an incident. This list holds 40 dated responses to AI risk where no harm was reported: banks restricting public chatbots such as ChatGPT, pausing a deployment, regulators issuing warnings, and internal probes. Each is sourced; none is counted in the incident register. The most recent is "RBC weighs blocking Meta's Muse AI agent; BMO CEO warns against giving agents banking credentials" (Oct 20, 2026).
These items are not AI incidents. They show how banks and regulators reacted to AI risk; the incident register is at /ai-incidents.
Oct 20, 2026 · Other · Royal Bank of Canada, BMO Financial Group
RBC weighs blocking Meta's Muse AI agent; BMO CEO warns against giving agents banking credentials
Royal Bank of Canada chief executive Dave McKay said on 20 October 2026 that the bank was considering blocking Meta's Muse AI agent from accessing customer information, citing concerns about monitoring transactions and data taken "without their consent or fraudulent consent", Bloomberg reported; he said no final decision had been made. The next day BMO chief executive Darryl White told the Toronto Global Forum that customers should not give AI agents their banking username, password, credit card or PIN, CBC News reported. Neither bank reported an incident.
Oct 14, 2026 · Other · Bank of America
Bank of America adds 1,000 contractors to patch software flaws exposed by Anthropic's Mythos
Bank of America chief executive Brian Moynihan told CNBC on 14 October 2026 that the bank put on 1,000 contractors in the second quarter, on top of 3,000 internal staff, to patch proprietary and vendor software vulnerabilities exposed by frontier models including Anthropic's Mythos. He told analysts the remediation and anti-money-laundering work cost "a couple hundred million plus" in the third quarter and expected the costs to decline, The Register reported. The work was defensive patching; no breach was reported.
Oct 13, 2026 · Other · JPMorgan Chase, Citigroup, Bank of America, Charles Schwab, Fifth Third
JPMorgan, Citi, Bank of America, Schwab and Fifth Third block or limit outside AI agents in customer accounts
In October 2026 several large US banks said they would not let third-party AI agents such as Meta's Muse log in or move money for customers. On 13 October Jamie Dimon told analysts JPMorgan would not allow "just any" outside agent, and a spokeswoman told CNBC the bank blocks third parties using customers' credentials. Schwab's CEO said on 15 October it blocks agents using client logins, American Banker reported. The Financial Times reported Citigroup blocks agents from logging in or transacting and Bank of America limits their access to deposits, according to PYMNTS and Banking Dive; Fifth Third's CFO said it does not let agents initiate transactions. No incident was disclosed.
Sep 30, 2026 · Regulator warning · Australian Securities and Investments Commission (ASIC)
ASIC to review Australian banks' AI use cases and their impact on customers
On 30 September 2026 ASIC published its 2026–27 supervisory letter to banks, saying it would commence a banking sector AI review "focused on new and proposed use cases of AI by banks and the impact on customers", expected to start in the second quarter of 2026–27 and coordinated with APRA. The letter recalls that ASIC's 2024 Report 798 found governance and risk assessment lagging AI adoption. Reuters, which reviewed the letter, reported ASIC wanted customers to remain protected in decision-making and lending processes. No bank was named and no finding has been made.
Sep 22, 2026 · Other · Bank of America, Capital One, NatWest Group, ING Group, Commonwealth Bank of Australia, ASB Bank
Six banks including Bank of America and NatWest warn AI shopping agents raise fraud and privacy risks
On 22 September 2026 ASB Bank, Bank of America, Capital One, Commonwealth Bank of Australia, ING Group and NatWest Group jointly published principles for "trusted agentic commerce", Bank of America said. Reuters, via CNBC, reported that the banks' paper warned AI shopping agents could increase the risk of scams, fraud and data-privacy breaches, including agents entering customers' card details directly into websites, and that the technology was advancing faster than standards and consumer protections. The paper was a warning; no incident was described.
Aug 31, 2026 · Regulator warning · Financial Supervisory Service (Korea)
Korea's FSS weighs review of financial call centres over AI chatbot complaints
South Korea's Financial Supervisory Service was reviewing plans to examine call-centre operations and complaint handling across banks, insurers, brokerages and card companies, the Seoul Economic Daily reported on 31 August 2026, citing industry sources, amid complaints that AI customer service delays connection to human agents and cannot resolve complex cases. An FSS official was quoted as saying the regulator planned "to review the complaint-handling systems at financial firms across the board"; News Tomato carried the same remarks on 3 September. No formal FSS announcement, named firm or finding had been published in these reports.
Jul 28, 2026 · Paused deployment · DBS Group
DBS holds off on letting AI agents act on their own while controls lag capability
DBS chief data and transformation officer Nimish Panchmatia told Computer Weekly in July 2026 that the bank would not let its AI agents act without a human checking their work, because governance of agents was innovating at "1x" against "5x" for capability: "we need to close this gap before we allow autonomy." DBS uses a chain of roughly 70 to 80 agents to assemble corporate credit memos, which relationship managers still review. He said agents taking unapproved shortcuts had been documented elsewhere in the industry, "though not at DBS". In an October 2025 DBS article he wrote that human-supervised governance "sits at the heart of all our AI deployments". No incident was described.
Jul 24, 2026 · Other · Itaú Unibanco
Itaú Unibanco alerts customers to deepfake and voice-clone scams
Itaú Unibanco warned customers in a release dated June 2026, posted on its investor-relations site in July 2026, that scams using deepfakes, voice cloning and other image, video and audio manipulation are gaining ground, including fake bank call centres and fake relatives or lawyers using cloned voices. Felipe Tambelini, its director of fraud prevention, said AI tools "tornaram os golpes mais sofisticados" (have made scams more sophisticated). The bank pointed customers to its Protected Mode and Pix Alert features and said it never asks for transfers, passwords or security codes by phone. This is an awareness notice; no specific incident or loss at Itaú was described.
Jun 12, 2026 · Other · UBS Group, KPMG
UBS says KPMG report's apparently AI-hallucinated claims about its AI agents were 'factually incorrect'
In June 2026 KPMG removed its October 2025 report "Redefining excellence in the age of agentic AI" after research group GPTZero told the Financial Times that inaccuracies in it stemmed from AI hallucinations, Finextra and TechCrunch reported. The report said UBS "integrates AI agents across investment advisory, risk management and compliance monitoring"; UBS told the FT this was "factually incorrect" and that it would ask KPMG to remove the claims. KPMG said it had removed the report while it investigated. The error was in a consultancy's publication, not in a UBS system, and no harm to UBS clients was reported.
May 21, 2026 · Regulator warning · Standard Chartered, Hong Kong Monetary Authority, Monetary Authority of Singapore
HKMA and MAS question Standard Chartered after CEO's 'lower-value human capital' AI remark
On 19 May 2026 Standard Chartered said it would cut about 15% of back-office roles by 2030, about 7,800 jobs, with chief executive Bill Winters citing automation and AI and describing it as "replacing in some cases lower-value human capital". Bloomberg reported on 21 May, as relayed by Reuters, that the Monetary Authority of Singapore discussed the remark with the bank and the Hong Kong Monetary Authority asked it to explain, including whether AI was a pretext for cuts; the HKMA said only that it "regularly engages" with banks. Winters apologised on 22 May for the upset his remark caused while standing by the explanation that lower-value roles are more vulnerable to automation.
May 12, 2026 · Regulator warning · Financial Services Agency (Japan), Mizuho Financial Group
Japan's FSA sets up public-private working group on Mythos-enabled cyber threats
Japan set up a public-private working group in May 2026 on cybersecurity risks to the financial system posed by Anthropic's Mythos model, Finance Minister Satsuki Katayama said on 12 May, Reuters reported. According to the Financial Services Agency, 36 entities including megabanks, internet banks, the Bank of Japan and the Japanese units of Anthropic and OpenAI would meet, chaired by Mizuho Financial Group's chief information security officer. Nikkei Asia reported on 27 August 2026 that top Japanese banks had regained access to Mythos after the US government ordered overseas access suspended in June. No incident was reported.
Apr 30, 2026 · Regulator warning · Australian Prudential Regulation Authority (APRA)
APRA warns banks and insurers of AI governance and AI-agent control gaps
On 30 April 2026 APRA wrote to all regulated banks, insurers and superannuation trustees after a late-2025 review of selected large entities, warning that governance, risk management, assurance and operational resilience "are not keeping pace" with AI adoption. The letter said many boards lacked the technical literacy to challenge AI risks, relied too heavily on vendor presentations, and that identity and access controls had not adjusted to non-human actors such as AI agents. APRA said it would take stronger supervisory action and, where appropriate, pursue enforcement where entities fail to control AI risks. No entity was named.
Apr 29, 2026 · Regulator warning · Office of the Superintendent of Financial Institutions (OSFI)
OSFI warns Canadian lenders that frontier AI such as Claude Mythos compresses cyber response times
Canada's banking regulator, the Office of the Superintendent of Financial Institutions, published a technology risk bulletin dated April 2026 saying frontier AI models such as Anthropic's Claude Mythos Preview "significantly compress the timeframe to respond to risks" and setting out sound practices on governance, patching and resilience. Reuters reported in July 2026, citing documents obtained under access-to-information law, that OSFI emailed the warning on 29 April to technology, security and risk chiefs at the largest banks and insurers. The warning was sector-wide and named no institution; no incident was reported.
Apr 28, 2026 · LLM restriction · Goldman Sachs
Goldman Sachs removes Anthropic Claude access for its Hong Kong bankers
Goldman Sachs removed its Hong Kong bankers' access to Anthropic's Claude on its internal AI platform in April 2026, according to a source cited by Reuters on 29 April after the Financial Times first reported it. The FT said Goldman took a strict interpretation of its contract with Anthropic and concluded Hong Kong staff should not use any Anthropic products; Gemini and ChatGPT remained available, Reuters' source said. Goldman declined to comment. An SCMP column in June 2026 said JPMorgan Chase later took the same step. No incident was reported.
Apr 20, 2026 · No-harm statement · Deutsche Bank
Deutsche Bank CEO says Mythos cyber risk is not causing panic at German banks
Deutsche Bank chief executive Christian Sewing said on 20 April 2026, speaking as head of the Association of German Banks, that Anthropic's Mythos model was "not something that leads to panic or alarm bells ringing" but had to be considered in daily risk management, Bloomberg reported. Reuters reported the same day that Sewing said banks were in close contact with European watchdogs about Mythos and that it was right that access to the model is limited for now. No incident was reported.
Apr 20, 2026 · Regulator warning · Monetary Authority of Singapore, DBS Group, OCBC
MAS urges Singapore banks to close cyber gaps over frontier AI; DBS says Mythos 'amplifies the risk'
On 20 April 2026 the Monetary Authority of Singapore said advances in AI could speed up the discovery and exploitation of software vulnerabilities and that financial institutions must "proactively identify and close vulnerabilities", Reuters reported. On 30 April DBS chief executive Tan Su Shan said Anthropic's Mythos "amplifies the risk" but that DBS still sees AI as a net positive, and OCBC said all such solutions undergo rigorous assessment before deployment, Reuters reported. The Business Times reported on 5 May 2026 that MAS had convened the chief executives of major financial institutions on AI-enabled cyber threats. No incident was reported.
Apr 17, 2026 · Other · Barclays
Barclays CEO warns Anthropic's Mythos is a serious threat to the global banking system
Barclays chief executive C. S. Venkatakrishnan said on 17 April 2026 that Anthropic's frontier AI model Mythos is a serious threat to the global banking system and is likely to be followed by more powerful successors, Reuters reported from a G30 meeting on the sidelines of the IMF spring meetings. "There will be a Mythos 2 and a Mythos 3," he said, adding that such models may be especially challenging for older institutions running legacy systems. The comments were a warning; no incident at Barclays was reported.
Apr 14, 2026 · Other · Ramp
Ramp fixes prompt-injection flaw that let its Sheets AI exfiltrate financial data
Security firm PromptArmor disclosed on 14 April 2026 that an indirect prompt injection hidden in white-on-white text in an imported dataset could make Ramp's agentic Sheets AI insert an IMAGE formula that sent confidential financial-model data to an attacker's server, with no user approval required. PromptArmor said it reported the flaw to Ramp on 19 February 2026 and that Ramp said on 16 March it was "resolved earlier today"; Gigazine relayed the findings. The demonstration used a test workbook; neither source reports exploitation of real customer data, and Ramp's response is known only through PromptArmor's disclosure timeline.
Apr 9, 2026 · Other · Nubank (Nu Holdings)
Deepfake ad campaign clones Nu founder David Vélez to push fake investments in Colombia
Forbes Colombia reported on 9 April 2026, using monitoring data from detection firm Revelum.ai, that between 10 March and 9 April 2026 a network ran 51 fraudulent Facebook and Instagram ads built on six unique deepfake videos of Nu co-founder and CEO David Vélez, promoting fake investment schemes. Nu had already warned in September 2025 that scams used Vélez's image in "deepfake" videos; Vélez said then that "ni Nu ni yo ofrecemos productos de inversión" (neither Nu nor I offer investment products) and that Nu was working with platforms and authorities. No victim losses tied to the campaign were reported. Listed as a response because the 2026 campaign figures come from a vendor via the press.
Mar 30, 2026 · Other · bunq
Dutch financial ombudsman finds bunq chatbot advice not at fault but upholds GDPR complaint
In binding ruling 2026-0302 of 30 March 2026, Kifid's Disputes Committee rejected a bunq customer's claim of €8,159.66 after she followed a bunq chatbot's suggestion in May 2025 to move her balance into a savings account and then hit its two-withdrawals-a-month limit, leaving her without full access to her money from 15 May to 1 August 2025. The committee found the chatbot had told her right away that she had two withdrawals left that month, and that using a chatbot is not automated decision-making under the GDPR. It upheld her complaint that bunq could not show it gave a timely substantive answer to her GDPR request. Ziptone reported the ruling on 10 April 2026.
Mar 11, 2026 · LLM restriction · Chinese state-owned banks
China tells state agencies and state-owned firms, including big banks, to curb the OpenClaw AI agent
Bloomberg reported on 11 March 2026 that Chinese government agencies and state-owned enterprises, "including the largest banks", had received notices warning them against installing the open-source OpenClaw AI agent on office devices for security reasons, according to people familiar with the matter; some staff were told to report existing installations for checks and possible removal. Reuters, citing two sources, said the scope of the curbs was unclear, and the South China Morning Post reported that several brokerages and banks had restricted staff access. No bank was named and no security breach was reported.
Feb 4, 2026 · Other · Commonwealth Bank of Australia
Commonwealth Bank says it may limit customers' use of external AI agents
In its report "Our Approach to Adopting AI", released in February 2026, Commonwealth Bank of Australia said customers may interact with its services through external AI agents outside its control and that, where it identifies material risk, it may implement limitations or controls to protect customers, channels and platforms. Capital Brief reported the statement on 5 February 2026 as CBA weighing limits on customers' use of external AI bots. The bank described no incident.
Aug 18, 2025 · Other · Standard Bank
Standard Bank warns of AI voice clones and deepfakes in scams impersonating its staff
On 18 August 2025 Standard Bank warned clients of a rise in "spoofing" scams in recent months in which fraudsters use AI-generated voices and emails to impersonate bank officials, with calls and emails that appear to come from the bank's legitimate contact details, TechCentral and BusinessTech reported. Its head of fraud risk management, Athaly Khan, said: "With the rapid development of artificial intelligence, we have seen an alarming enhancement in spoofing techniques." She named voice cloning, deepfake videos, chatbots and AI-generated phishing emails. The bank gave no count of cases or losses; this is a customer warning, not a dated incident.
Jul 14, 2025 · Other · PayPal
PayPal warns of AI-written fake-purchase emails leading to tech-support scams
In July 2025 PayPal issued consumer alerts about phishing emails made more believable with artificial intelligence, Pennsylvania broadcaster WGAL reported on 14 July; eSecurityPlanet relayed the warning the next day. The emails, addressed to "Dear PayPal Customer", claim purchases such as a $700 iPhone or $599 of Bitcoin and give a phone number; WGAL said callers were asked to download a PayPal-branded "support tool" that gives scammers full access to their device. It was a general warning: no campaign size, victim count or loss figure was published.
Jun 10, 2025 · LLM restriction · Goldman Sachs
Goldman Sachs tells campus candidates ChatGPT is prohibited during interviews
In June 2025 Goldman Sachs' EMEA campus recruitment team emailed students reminding them that the bank prohibits external sources, including ChatGPT and Google search, during its interview process, which uses the HireVue video platform, eFinancialCareers and Fortune reported. A Goldman spokesperson told Fortune the language is consistent with what the bank sends to all campus applicants: "We want to hear from our applicants in their own voice." The notice was a policy reminder; no incident was reported.
Nov 13, 2024 · Regulator warning · FinCEN
FinCEN alerts financial institutions to deepfake media fraud
On 13 November 2024 the US Financial Crimes Enforcement Network issued Alert FIN-2024-Alert004 on fraud schemes using deepfake media created with generative AI. FinCEN said that, beginning in 2023 and continuing in 2024, it had observed an increase in suspicious activity reports describing suspected deepfakes, often altered or fabricated identity documents used to get past identity verification. The alert lists red flags, such as a customer using a third-party webcam plugin during a live check, and asks institutions to tag related reports "FIN-2024-DEEPFAKEFRAUD". It names no bank and reports no loss figure.
Oct 29, 2024 · Regulator warning · Australian Securities and Investments Commission (ASIC)
ASIC warns licensees' AI governance is lagging adoption (REP 798)
In October 2024 ASIC published Report 798, "Beware the gap", after analysing 624 AI use cases at 23 licensees in banking, credit, insurance and financial advice as at December 2023. The regulator found AI use so far was mostly cautious but that governance and risk management did not always keep pace, and in some cases lagged adoption, "creating the greatest risk of consumer harm". ifa reported that nearly half of the licensees had no policies addressing consumer fairness or bias. Chair Joe Longo said work was needed "quickly". The report did not name the licensees and made no enforcement finding.
Sep 18, 2024 · Other · Starling Bank
Starling Bank warns millions could be caught out by AI voice-cloning scams
On 18 September 2024 Starling Bank launched a "Safe Phrases" campaign, backed by the UK government's Stop! Think Fraud campaign and fronted by actor James Nesbitt, warning that AI voice-cloning scams could "catch millions out", CNN and The Guardian reported. A Mortar Research survey of more than 3,000 UK adults for the bank found 28% said they had been targeted by an AI voice-cloning scam in the past year and 46% did not know such scams existed. Starling's chief information security officer Lisa Grahame said scammers "only need three seconds of audio" to clone a voice. This is an awareness campaign; no specific incident or loss was described.
Aug 7, 2024 · Other · Santander UK
Santander UK publishes deepfakes of its own fraud lead to warn customers about AI scams
On 7 August 2024 Santander UK said it had created deepfake videos of its head of fraud risk management, Chris Ainsley, and of "finfluencer" Mr Money Jar and posted them on social media to show how realistic AI deepfakes already are. The bank's own research found 53% of Britons had not heard of the term deepfake or misunderstood it, and only 17% were confident they could identify a deepfake video. Ainsley said it was "'when' rather than 'if'" deepfake-driven scams would surge. This was an awareness campaign; the bank described no specific incident or loss.
Nov 17, 2023 · Other · Upstart, US Securities and Exchange Commission
SEC subpoenas Upstart over disclosures on its AI models and loans
Upstart disclosed in May 2024 that on 17 November 2023 it had received a subpoena from the SEC seeking documents and information about its disclosures, "including the use of our AI models and loans." The lender said it was cooperating and could not predict the outcome, and told Banking Dive it had nothing to add beyond its filing. The same wording appears in its annual report for 2024. Its 10-Q filings for the quarters ended 30 June 2025 and 30 June 2026 do not mention the subpoena, and neither the SEC nor Upstart has announced any action or finding.
Jun 6, 2023 · Regulator warning · Bank of America, Capital One, Wells Fargo, Consumer Financial Protection Bureau
CFPB warns banks that poorly deployed chatbots can breach consumer law
The CFPB's June 2023 issue spotlight on chatbots in consumer finance said all of the top ten US commercial banks use chatbots and that it had received numerous complaints from customers stuck in "doom loops" without a route to a human. It described Capital One's Eno, Bank of America's Erica and Wells Fargo's Fargo, and warned that inaccurate chatbot answers or a failure to recognise a consumer invoking federal rights risk breaching consumer financial law. Then-Director Rohit Chopra said a poorly deployed chatbot can lead to "even violations of the law." The spotlight made no finding against any named bank.
May 24, 2023 · Other · Binance
Binance says deepfakes are increasingly used to try to spoof its identity checks
Binance chief security officer Jimmy Su told Cointelegraph in May 2023 that there had been a rise in fraudsters using AI deepfake videos, built from photos of victims found online, to try to get past the exchange's customer verification. He said the fakes could already follow liveness prompts such as blinking or turning the head, but that human reviewers could still spot flaws, adding that "AI will overcome [them] over time". Gadgets360 relayed the interview. Binance disclosed no number of successful bypasses and no loss figure.
May 4, 2023 · Regulator warning · JPMorgan Chase, Bank of America, Wells Fargo, Morgan Stanley, Charles Schwab, TD Bank
Senate Banking chair presses six banks on voice authentication after AI voice clones fooled bank systems
On 4 May 2023 Senate Banking Committee chair Sherrod Brown wrote to the chief executives of JPMorgan Chase, Bank of America, Wells Fargo, Morgan Stanley, Charles Schwab and TD Bank about their voice authentication services, citing Vice and Wall Street Journal reporters who had used AI-generated voice clips to fool bank voice authentication. The letters asked how the banks secure the tools and protect customers' voice data, with responses requested by 18 May 2023. The letters were an inquiry; they did not allege losses at the six banks.
Apr 10, 2023 · Regulator warning · Payment & Clearing Association of China
PBoC-governed payment association tells payment staff not to upload confidential data to ChatGPT
On 10 April 2023 the Payment & Clearing Association of China, an industry body governed by the People's Bank of China, warned against using ChatGPT and similar AI tools because of "risks such as cross-border data leaks", Reuters reported in coverage carried by The Times of India and summarised by Competition Policy International. The association said payment-industry staff must comply with laws and rules when using such tools and "should not upload confidential information related to the country and the finance industry." No data leak at a named institution was reported.
Feb 22, 2023 · LLM restriction · JPMorgan Chase, Bank of America, Citigroup, Goldman Sachs, Wells Fargo, Deutsche Bank
JPMorgan, Bank of America, Citi, Goldman Sachs, Deutsche Bank and Wells Fargo restrict staff use of ChatGPT
In February 2023 six large banks restricted staff use of OpenAI's ChatGPT. JPMorgan Chase curbed it firm-wide, Bloomberg and CNN reported on 22 February, citing a person familiar who said no specific incident triggered the move; the bank later confirmed the restriction to HR Brew (11 May 2023). On 24 February Bloomberg reported that Bank of America, Citigroup, Deutsche Bank, Goldman Sachs and Wells Fargo had also banned its use, with Bank of America telling staff it was prohibited for business use. Forbes reported the same day that Citi and Goldman restricted it through automatic limits on third-party software and that Wells Fargo confirmed it was "imposing usage limits". Deutsche Bank told HR Brew it had blocked the website as "protection against data leakage". No data leak was reported.
Jan 26, 2023 · Other · ING Bank România
ING Bank Romania and Romania's cyber directorate warn that scammers use AI to write phishing
On 26 January 2023 ING Bank România and Romania's National Cyber Security Directorate (DNSC) jointly warned card holders that online fraud attempts would become more elaborate and harder to detect because attackers had begun using artificial-intelligence tools, HotNews.ro and Wall-Street.ro reported. ING's head of information security, Bogdan Costea, said current AI applications "pot genera un mesaj de phishing care pare autentic" (can generate a phishing message that looks authentic), without grammatical errors, including in Romanian. The warning described a trend, not a specific breach; no losses at ING were reported.
Oct 18, 2022 · Other · bunq
Dutch appeals tribunal finds DNB failed to prove bunq's AI-based customer screening unlawful
On 18 October 2022 the Dutch Trade and Industry Appeals Tribunal (CBb) ruled that De Nederlandsche Bank had not proven that bunq's screening of retail customers, built on a standard user profile developed with data analysis and artificial intelligence, breached anti-money-laundering law, and partly struck down DNB's instructions to the bank. The court said bunq had explained at length how its AI-based method worked and DNB had not countered it sufficiently. The CBb did find that bunq broke other AML rules, including source-of-funds checks and enhanced due diligence on politically exposed persons. Pinsent Masons' Out-Law reported the ruling on 26 October 2022.
Jun 8, 2022 · Other · Upstart, Consumer Financial Protection Bureau
CFPB terminates Upstart's AI-lending no-action letter at Upstart's request
On 8 June 2022 the CFPB issued an order terminating Upstart's 2020 no-action letter, which had shielded the lender's "artificial intelligence" underwriting model from fair-lending charges, effective immediately. The CFPB said Upstart notified it on 13 April 2022 that it would add a significant number of new variables to its model; the bureau needed time to review them, and Upstart asked to end the letter so it could change the model without CFPB review. Upstart said the request was motivated by keeping its risk models accurate during a period of economic change. The order made no finding of wrongdoing.
May 26, 2022 · Regulator warning · Consumer Financial Protection Bureau
CFPB: complex algorithms do not excuse vague credit-denial reasons (circulars later withdrawn)
On 26 May 2022 the CFPB published Circular 2022-03, saying the Equal Credit Opportunity Act requires lenders to give specific reasons for denying credit even when they rely on complex "black-box" algorithms. On 19 September 2023 Circular 2023-03 added that lenders using AI cannot simply pick the closest reasons from the CFPB's sample checklists; then-Director Rohit Chopra said "there is no special exemption for artificial intelligence." Neither circular named a lender. The CFPB's withdrawn-guidance page lists both circulars as withdrawn on 12 May 2025, though ECOA's notice requirement itself was not repealed.
Mar 29, 2022 · Other · Kasikornbank (KBank)
KBank warns of call-centre gangs using deepfake police officers on video calls
On 29 March 2022 Kasikornbank (KBank) managing director Krit Jitjang warned that call-centre scam gangs had adopted a new trick: a video call with a fake police officer generated with deepfake technology that moved only the mouth and added a threatening voice, Thai PBS and MGR Online reported. According to the bank's account, a KBank customer told that a parcel contained a passbook linked to money laundering transferred more than 600,000 baht before realising it was a scam. KBank said it was working with police, flagging unusual accounts and running its "use SATI to protect SATANG" awareness campaign. The bank's own release appears only on a PR wire; no recovery figure was reported.
When AI goes wrong at a bank, the next morning's brief says so.
when one of these regulators moves, the next morning's brief says so · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning