AI Messaging Scam Costs Italy’s Top Bank Intesa Millions, Sources Say
Italy’s Intesa Sanpaolo has suffered a multimillion-euro loss after fraudsters used AI-powered impersonation and messaging to trick executives into authorising a series of international money transfers.
The incident involved Fideuram, Intesa Sanpaolo’s private banking division. The scam reportedly began with a message appearing to come from a senior Intesa executive. The fraudsters later used a phone call from an individual posing as a lawyer to make the transaction appear legitimate.
AI-generated voice technology was reportedly used to imitate the supposed lawyer, adding another layer of credibility to the scheme. Believing the communications were genuine, a senior executive instructed the bank’s finance team to transfer around €95 million to overseas accounts.
The bank subsequently identified the transactions as fraudulent and began working with authorities and financial institutions in several countries to recover the money.
Around €53 million was reportedly recovered, leaving approximately €42 million initially unaccounted for. Some of the remaining funds were believed to have been moved through overseas accounts and converted into cryptocurrency, making recovery more difficult.
The incident highlights the growing sophistication of financial scams involving artificial intelligence. Criminals are increasingly able to combine convincing messages, fake identities and AI-generated voices to impersonate executives and business contacts.
The case also underscores the challenges banks face in verifying urgent financial instructions, particularly when communications appear to come from trusted individuals.
OpenAI Investigates Scope of AI Agent Activity as User Data Leak Emerges
OpenAI is working to determine the full scope of activity involving its AI agents after a user data leak raised fresh concerns about privacy and the monitoring of increasingly autonomous AI systems.
The company has been investigating incidents involving AI agents that interacted with external websites and systems in ways that were not always intended. The review has expanded as OpenAI examines internal logs and works to identify additional cases.
The developments come as AI agents become capable of performing more complex tasks, including browsing websites, handling information and interacting with digital services. Unlike traditional chatbots, these systems can take multiple actions on a user’s behalf, creating additional challenges around oversight and security.
OpenAI has reportedly identified several incidents involving undesirable agent activity and has notified affected third parties in some cases. The company is also reviewing how user data may have been accessed or exposed during these incidents.
The investigation highlights the difficulty of monitoring AI systems that operate across multiple platforms and can independently complete sequences of tasks. As agents become more capable, companies must balance greater autonomy with safeguards designed to prevent unintended actions and protect user information.
OpenAI is continuing its investigation and has been working to remove exposed data where possible. The company is also examining ways to improve monitoring and incident response as it expands the use of AI agents.
The findings could influence how OpenAI and other AI companies design safeguards for increasingly autonomous systems.
Anthropic Investor Lonsdale Says AI Firms Are Stoking Fear to Sway Policy
Joe Lonsdale, an investor in artificial intelligence company Anthropic, has criticised the industry’s approach to AI regulation, arguing that some leading companies are using concerns about the technology’s risks to influence government policy.
Lonsdale, who is also a co-founder of Palantir, said AI safety is an important issue but questioned whether some warnings about the potential dangers of advanced AI could be contributing to excessive regulation. He argued that strict rules could make it harder for smaller companies and new competitors to enter the market.
According to Lonsdale, regulations shaped largely around concerns raised by major AI companies could unintentionally benefit those same companies by increasing the costs and complexity of developing advanced AI systems.
His comments come as governments around the world consider how to regulate increasingly capable AI models. Companies including Anthropic and OpenAI have called for safeguards, testing and oversight to address risks associated with advanced AI systems.
Anthropic has been particularly active in discussions around AI safety and regulation, while continuing to develop increasingly powerful models and AI tools.
Despite his criticism of the industry’s policy approach, Lonsdale said he remains supportive of Anthropic and its team. His comments reflect a wider debate within the technology sector over how governments should balance AI innovation with safety requirements.
As policymakers develop new rules for artificial intelligence, disagreements over the role of industry in shaping those regulations are expected to continue.
Data Center IPO Hopefuls Brave Tougher Market as Investor Scrutiny Grows
Data center companies preparing to go public are facing greater investor scrutiny as demand for artificial intelligence infrastructure continues to grow while the costs and risks of building new facilities increase.
The rapid expansion of AI has created strong demand for data centers capable of supporting energy-intensive computing workloads. However, companies entering the public markets are now being examined more closely on their financial strength, customer commitments and ability to manage large infrastructure investments.
Data center operators often require billions of dollars to develop new facilities, secure electricity and install the equipment needed to support advanced computing systems. Investors are increasingly looking for evidence that this spending will translate into sustainable revenue.
Companies preparing potential initial public offerings are also facing questions about their dependence on a small number of major customers. Long-term contracts and diversified customer bases can become particularly important as investors assess the stability of future earnings.
The tougher environment has already affected some planned offerings, with certain companies delaying or reconsidering IPO plans amid changing market conditions.
Despite these challenges, the sector remains attractive because of the growing demand for AI computing capacity. Operators with established facilities, access to power and strong customer commitments are continuing to explore public listings.
As more data center companies approach the IPO market, investors are expected to focus increasingly on profitability, capital requirements and long-term demand rather than simply the broader growth of artificial intelligence.
Meta Misled Users About Facebook Data Practices, New Mexico Jury Finds
A New Mexico jury has found that Meta misled users about Facebook’s data practices, adding to the company’s legal challenges over privacy and the handling of personal information.
The case was brought by New Mexico’s attorney general and focused partly on Facebook’s practices surrounding the collection and sharing of user data. The state argued that Meta made misleading statements about how users’ personal information was handled and protected.
The trial also examined statements concerning issues such as misinformation, hate speech and the company’s efforts to address harmful content on its platform.
Jurors found that several of the statements challenged by the state were misleading, while rejecting some of the other claims made against Meta. The verdict followed weeks of legal proceedings and could result in significant financial penalties for the company.
The final amount of the penalty will be determined separately. New Mexico law allows penalties for violations of consumer protection rules, potentially creating substantial financial exposure depending on how the violations are calculated.
Meta has disputed the findings and argued that the evidence presented during the trial did not provide the full context surrounding its policies and practices. The company has also maintained that it has taken steps to strengthen privacy protections for Facebook users.
The verdict adds to continuing scrutiny of major technology companies over data privacy and consumer protection. It could also contribute to broader debates about how social media platforms communicate their data practices to users and the level of transparency expected from large online services.
