This week’s stories share one theme: attackers getting in through channels people already trust. Prosecutors described AI-cloned executive voices that cost a Missouri hospital system nearly $5 million. A record crypto theft began with a job interview. A lending platform used by community banks went dark. Federal agencies warned about attacks on everyday refrigeration equipment. Stolen AI keys are producing six-figure cloud bills. And Google revised a plan that tied school access to Search to its AI tools.
Three charged in $129 million fraud that used AI to impersonate executives; a Missouri hospital system lost $4.8 million
Federal prosecutors in Manhattan have charged three men with using artificial intelligence to impersonate corporate executives and attorneys at large law firms. According to the indictment, the scheme ran from July 2024 to February 2026 and cost companies more than $129 million. ABC News reports that the victims included an unnamed hospital system headquartered in Missouri, which lost $4,829,580, along with a California hospital system, a Rhode Island bank, and a Florida pet food company.
According to the Associated Press, the scheme usually began with an email that appeared to come from the CEO and asked accounting staff for help with a “confidential financial transaction.” Phone or video calls followed, in which AI mimicked the executive’s voice and sometimes appearance. Employees were then handed off to a fake attorney, who gave the wiring instructions. AP reports the stolen funds were converted to cryptocurrency.
All three defendants were arrested in Europe and are awaiting extradition. These are allegations in an indictment, not convictions. U.S. Attorney Jamie McDonald said the public “should recognize that there are some who seek to use this emerging technology to pursue criminal goals.”
Why it matters
- A convincing voice or video call is no longer proof of identity.
- A Missouri healthcare organization lost nearly $5 million, which makes this a local concern, not a distant one.
What organizations should consider
- Require verification of any urgent or confidential wire request through a separately known phone number or in person, never through contact details supplied in the request itself.
- Use dual approval for large or unusual payments, so no single employee can be pressured into sending funds.
- Tell finance staff about this specific pattern: a CEO request for confidentiality, followed by a hand-off to an outside “attorney.”
Original article: ABC News Supporting coverage: SecurityWeek (Associated Press)
Binance says a fake job interview helped hackers steal $6.3 billion
Binance has described how attackers carried out the record $6.3 billion theft from the cryptocurrency exchange, Nextgov/FCW reports. The company traces the campaign to June, when an employee downloaded a tampered software installer while setting up an AI coding tool. Binance caught that intrusion and replaced the computer, but by then the intruders had learned who worked on the company’s wallet systems.
In September, someone posing as a recruiter approached a different Binance engineer through a legitimate job site. After two interviews, the engineer was asked to download a malicious assignment, which ran on his work computer. On October 20, the attackers used that computer and an internal Binance AI tool to take control of two wallet engineers’ computers. Nine days later, they used one engineer’s credentials on the other’s machine to reach the service that approves transfers.
The FBI attributed the theft to North Korea, which has denied involvement, according to state media cited by Yonhap. Binance said about $2.85 billion had been frozen or intercepted, and that it would cover customer losses. “This should not have happened,” the company said.
Why it matters
- Two routine actions, a software download and a job interview, gave attackers their way in.
- Employees who interview elsewhere on work devices create risk that most security programs do not account for.
- The attackers spent months learning who held the most sensitive access before they moved.
What organizations should consider
- Control what software employees can download and run on company computers, as security commentators quoted in the coverage recommend.
- Limit user permissions so that one compromised machine cannot reach critical systems.
- Pay closer attention to the computers of employees with access to payments, finances, or administrative systems.
Original article: Nextgov/FCW
Lending platform used by thousands of community banks and credit unions confirms data security incident
Abrigo, whose Sageworks platform is used by community banks and credit unions to process and monitor loans, has confirmed that a weekend outage was caused by a “data security incident,” American Banker reports. Abrigo disabled Sageworks on Friday, October 30, after noticing “irregular activity” in its environment. BleepingComputer reports that Abrigo says it serves more than 2,400 institutions, mainly in the United States.
Most of Sageworks was back online on Monday, November 2. Borrower-facing features, including the online loan application portal, remained down as of the American Banker report. Sageworks holds the documents behind commercial and small-business loans, including tax returns, financial statements, rent rolls, and credit reports. Abrigo said “the unauthorized access has been contained” but has not said whether any data was accessed or taken.
American Banker describes this as Abrigo’s third security incident in about 14 months, following a 2025 intrusion into its Salesforce account and a 2026 extortion attempt involving data allegedly taken from that account.
Why it matters
- Many Missouri and Midwest businesses borrow from community banks and credit unions, and their loan files contain detailed financial records.
- A single vendor outage stalled lending work at many institutions at once.
- Whether borrower data was exposed remains unknown.
What organizations should consider
- Businesses with recent or pending loan applications can ask their lender whether it uses the affected platform and how it is communicating with borrowers.
- Financial institutions using the platform can follow researchers’ advice, reported by BleepingComputer, to rotate Sageworks API keys.
- Review how much sensitive documentation critical vendors hold, and what notification terms apply if one has an incident.
Original article: American Banker Supporting coverage: BleepingComputer
U.S. agencies warn of ongoing attacks on refrigeration and cold chain systems
CISA, the FBI, and other federal departments issued a joint warning on Thursday about hackers compromising internet-connected refrigeration and cold chain monitoring systems, TechCrunch reports. The warning follows two waves of attacks in Europe that raised freezer and chiller temperatures in at least a dozen countries, spoiling food. Some of the affected systems were remote-controlled refrigeration units made by Danfoss. Hackers also targeted a blood bank in France, but the attack was stopped before blood supplies were compromised.
Citing Reuters, TechCrunch reports that the U.S. Department of Defense is investigating a series of hacks targeting freezers run by the military. The U.S. government has not attributed the attacks to a particular group or government. Security researchers have broadly linked them to Iran, which remains unconfirmed.
“Every organization that depends on refrigeration and monitoring systems should assume it could be a target,” said Chris Butera, CISA’s acting cybersecurity chief.
Why it matters
- Grocers, restaurants, clinics, pharmacies, food processors, and cold storage facilities rely on this equipment every day.
- Losses here mean spoiled inventory and disrupted operations, not stolen data.
- These systems are often installed and managed outside the IT department, so they can be overlooked.
What organizations should consider
- Remove remote control systems from the internet, as the agencies advise.
- Change and strengthen passwords and apply patches where possible.
- Retire equipment that is no longer supported by the vendor.
Original article: TechCrunch
Stolen AI access is leaving companies with cloud bills of up to $1 million
Google’s Threat Intelligence Group and Mandiant have described a growing practice called “LLMjacking,” in which criminals steal access to an organization’s cloud AI services and resell it, BleepingComputer reports. The victim’s cloud provider records the activity as legitimate use and bills the victim.
According to the research, initial access comes from exposed or stolen cloud credentials, service account keys, developer tokens, and AI service API keys. Some organizations saw fraudulent charges of more than $100,000 in a single day, and cumulative charges at some reached $1 million. In one case, a reseller passed nearly $1 million in usage costs to a single victim in under eight hours. The findings come from Google’s own research teams.
Why it matters
- As organizations experiment with AI tools, API keys are often created quickly and stored carelessly.
- The first sign of compromise may be a cloud invoice, not a security alert.
- Unapproved “shadow AI” use makes these keys harder to track.
What organizations should consider
- Store and rotate keys and secrets through a centralized management platform.
- Deny permissions that are not explicitly required, including access to cloud AI services.
- Enforce phishing-resistant multifactor authentication.
- Set API usage and spending limits, and monitor cloud audit logs for unusual activity.
Original article: BleepingComputer
Google revises plan that would have cut off Search for schools restricting its AI tools
Google told schools using its K-12 software that students would not be able to use Search unless the school opted in to Google’s AI tools, TechCrunch reports. Chris Phillips, Google’s vice president for education, told The New York Times that “Search and AI are deeply integrated.”
The plan drew criticism from Sen. Bernie Sanders, Sen. Richard Blumenthal, former FTC chair Lina Khan, New York City Mayor Zohran Mamdani, and educators. Google then said that Gemini, Gemini Notebook, and Search will be off by default, and that schools can turn each on separately. “We know there’s concern about students using AI to shortcut schoolwork, and schools have told us they want control,” Google said.
Why it matters
- School districts and other organizations increasingly find AI bundled into tools they already depend on.
- Changes to vendor defaults can quickly affect policy decisions an organization thought it had already made.
- The episode shows that public pushback can change how vendors roll out AI features.
What organizations should consider
- Review the AI settings and defaults in core platforms, and confirm they match internal policy.
- Assign someone to track vendor announcements that change how AI features are enabled or bundled.
- Make sure your acceptable-use policy says which AI tools are approved and for what purposes.
Original article: TechCrunch