OpenAI is releasing a new version of its flagship AI model designed to handle complex knowledge work, research tasks and software development more effectively. OpenAI CEO Sam Altman announces on X that GPT-5.4 is now available through the company’s API and Codex developer platform, with the update rolling out to ChatGPT users throughout the day. According to Altman, the new model improves several core capabilities used by professionals and developers. “GPT-5.4 is launching, available now in the API and Codex and rolling out over the course of the day in ChatGPT. It’s much better at knowledge work and web search,…
Author: CapitalAI Daily Team
Tax scammers are siphoning money from millions of Americans, with victims losing an average of $1,020 during the filing season, according to new research from McAfee. The cybersecurity firm says nearly 20% of Americans have lost money to a tax-related scam as criminals exploit confusion around tax rules and heightened anxiety during filing season. Abhishek Karnik, head of threat research at McAfee, says scams are now harder to detect with the rise of AI. “Tax season has long been prime time for scammers because it combines high stress with people sharing sensitive personal information. What’s changed is how real these…
Arizona recovered millions of dollars from scams last year, but state officials warn that the threat is becoming more sophisticated. Attorney General Kris Mayes announces that her office recovered more than $4 million in 2025 for residents harmed by scams, fraud and deceptive business practices. “My office fights every day to protect Arizonans from fraud and deceptive practices, and recovering more than $4 million for consumers last year is just part of the work we are doing to help Arizonans. But the threat is evolving and scammers are now using artificial intelligence to clone voices, create fake videos, and craft…
Apple is weighing deeper ties with Google even as questions mount over demand for its in-house AI tools. The Information Apple correspondent Aaron Tilley says the company signed a deal earlier this year to use Google’s Gemini model to power an upgraded Siri. According to Tilley, the deal could now extend beyond models and into broader cloud infrastructure. “In January, they signed this Gemini deal to use Google’s AI model to power upgraded Siri. A part of that is also a broader sort of cloud arrangement, where Apple is now exploring various ways to use more of Google’s cloud, and…
The U.S. Postal Inspection Service says artificial intelligence is supercharging traditional scams. In a new public warning, the agency says criminals are using AI-generated images and cloned voices to make scams appear more convincing. “Scammers are using AI-generated photos and voice clones that mimic a real person to build relationships with unsuspecting victims to ultimately get money and steal personal information. AI content can make romance and investment scams, cryptocurrency scams, scams claiming to be tech support, and even fake emergency requests from friends and family feel legitimate and believable.” Chief Postal Inspector Gary Barksdale says the tools now available…
Crypto exchange Bybit says it stopped hundreds of millions of dollars in suspected scam withdrawals using a new AI-based security system. In a new press release outlining its 2025 Security Initiative, the Dubai-based exchange says it intercepted and recovered $300 million out of $500 million in flagged withdrawals in the fourth quarter alone, protecting more than 4,000 users. Bybit says the system is built around a three-tier risk framework designed to intervene before funds leave the platform. At the lowest level, the system uses data analysis to spot unusual patterns, such as many withdrawals going to the same new wallet…
Amazon’s potential multi-billion-dollar investment in the ChatGPT creator OpenAI may depend on whether the AI company reaches a key technological milestone or goes public. Amazon is in talks to invest up to $50 billion in OpenAI under terms that are still being negotiated, according to Reuters, citing The Information. The structure of the deal includes $15 billion upfront, with an additional $35 billion contingent on OpenAI either achieving an artificial general intelligence (AGI) milestone or pursuing an initial public offering. SoftBank and Nvidia each plan to invest $30 billion in three installments over the year as part of the same…
OpenAI says it has shut down a network of accounts that used ChatGPT to run a fake “scam recovery” operation targeting fraud victims. In its latest threat disruption report, the company says it banned a cluster of accounts that posed as fictitious law firms and impersonated real attorneys and US authorities, including the FBI’s Internet Crime Complaint Center. OpenAI says the thieves used ChatGPT to generate promotional materials for at least six fake law firms. The content was posted through social media accounts and online ads that promoted fraudulent financial recovery services. According to the firm, the scammers tried to…
The White House is pushing major tech firms to generate their own power as AI data center expansion accelerates. Amazon, Google, Meta, Microsoft, xAI, Oracle and OpenAI are set to sign President Donald Trump’s “Rate Payer Protection Pledge” at the White House on March 4th, Fox News reports. Under the pledge, participating companies will agree to source their own power supply for new AI data centers rather than rely on the existing grid. White House spokeswoman Taylor Rogers says that the initiative is designed to prevent electricity costs from rising for American households as AI infrastructure scales. “Under this bold…
OpenAI says it shut down a ChatGPT account linked to a person associated with Chinese law enforcement after it found signs of a coordinated influence operation. In a new threat disruption update, OpenAI says the user’s activity pointed to what it described as a well-resourced effort to run covert influence operations against domestic and foreign targets. OpenAI says the account attempted to use ChatGPT to help plan a covert influence operation aimed at Japan’s prime minister, but the model refused. The company says the same account also used ChatGPT to edit periodic status reports about the broader campaign, which the…