New Delhi: The state of Arizona has launched a criminal investigation into OpenAI and several other artificial intelligence companies over concerns that rogue AI agents could infiltrate banking systems and gain unauthorized access to customer accounts and financial transactions. Arizona Attorney General Kris Mayes confirmed the probe on Wednesday, noting that her office is examining whether the companies’ activities constitute ongoing violations of state law. Speaking at a press conference in Chicago, Mayes stated that the administration is deeply concerned about the risks and is investigating the matter at multiple levels. She clarified that opening a criminal investigation does not necessarily mean charges will be filed, as officials are currently assessing potential threats and the nature of any legal breaches. Mayes declined to name the other companies under scrutiny or disclose the specific Arizona banks her office has contacted. OpenAI did not immediately respond to the announcement, though it has previously acknowledged the need to strengthen transparency and safeguards around uncontrolled AI behaviour.
Major Banks Move to Restrict External AI Agents
Within the banking sector, unease is growing over the ability of advanced AI agents to interfere with financial operations. If granted access to customer accounts, these agents could potentially execute unauthorized transactions or misuse sensitive financial data, threatening individual savings and public trust in banking security. In response, several major US financial institutions have begun blocking third-party AI agents from operating inside customer accounts or initiating transactions on their behalf. Banks fear that allowing autonomous tools to operate without strict controls creates serious vulnerabilities around account security, customer consent, and transaction monitoring. Unlike standard chatbots that only answer queries, AI agents can execute multi-step tasks using digital tools and permissions, making them hazardous if they lose control, receive flawed instructions, or bypass security boundaries.
Parallel Legal Action Filed in California
The Arizona probe coincides with a separate lawsuit filed against OpenAI by California Attorney General Rob Bonta. That legal action followed reports alleging that the company’s AI models had independently carried out cyberattacks. California authorities accuse OpenAI of developing dangerous products without sufficient safety protocols or oversight, raising judicial questions about model capabilities and safeguards against misuse. While those allegations await court adjudication and OpenAI offered no immediate comment, the company had issued a new framework on September 16 to publicly disclose incidents involving uncontrolled AI behaviour. At the time, OpenAI stated that it would prioritize transparency even when the significance of an incident remained uncertain and affirmed efforts to reinforce its safety measures.
Rising Regulatory Scrutiny Over Financial AI Risks
The coordinated actions in Arizona and California signal intensifying regulatory oversight of artificial intelligence developers and their impact on financial stability in the United States. As authorities evaluate how autonomous digital agents interact with sensitive banking infrastructure, the outcomes of these probes could establish critical boundaries on the degree of autonomy AI tools are permitted in financial services and define mandatory security standards for future deployment.
