The Federal Reserve Bank of New York and the Bank for International Settlements (BIS) have published a joint research study that explored how central banks could continue to implement monetary policy operations in tokenised wholesale financial markets. Dubbed Project Pine, the study found that central banks could deploy policy implementation tools using programmable smart contracts in a potential future state where commercial banks have widely adopted tokenisation for wholesale payments and securities settlement. The project generated the prototype of a generic monetary policy implementation tokenised toolkit for potential further research and development by central banks across jurisdictions and currencies. The BIS and the Fed say the prototype can fulfil a common set of central bank implementation requirements, including paying interest on reserves, open market operations, and collateral management. The toolkit was tested against ten hypothetical scenarios that applied historical data inputs on past market events, such as interest rate tightening and easing cycles, quantitative easing and tightening cycles, and periods of strained market liquidity or broader market disruptions. “The prototype successfully responded and instantaneously carried out the intended operation under the varying market conditions,” states the BIS. “Project Pine’s findings highlighted areas for further research and analysis related to interoperability and data standardisation.”
DarkBench is the first benchmark designed specifically to detect and categorize LLM dark patterns, AI sycophancy, brand bias or emotional mirroring
Esben Kran, founder of AI safety research firm Apart Research, and his team approach large language models (LLMs) much like psychologists studying human behavior. Their early “black box psychology” projects analyzed models as if they were human subjects, identifying recurring traits and tendencies in their interactions with users. “We saw that there were very clear indications that models could be analyzed in this frame, and it was very valuable to do so, because you end up getting a lot of valid feedback from how they behave towards users,” said Kran. Among the most alarming: sycophancy and what the researchers now call LLM dark patterns. Kran describes the ChatGPT-4o incident as an early warning. As AI developers chase profit and user engagement, they may be incentivized to introduce or tolerate behaviors like sycophancy, brand bias or emotional mirroring—features that make chatbots more persuasive and more manipulative. To combat the threat of manipulative AIs, Kran and a collective of AI safety researchers have developed DarkBench, the first benchmark designed specifically to detect and categorize LLM dark patterns. Their research uncovered a range of manipulative and untruthful behaviors across the following six categories: Brand Bias, User Retention, Sycophancy, Anthropomorphism, Harmful Content Generation, and Sneaking. On average, the researchers found the Claude 3 family the safest for users to interact with. And interestingly—despite its recent disastrous update—GPT-4o exhibited the lowest rate of sycophancy. This underscores how model behavior can shift dramatically even between minor updates, a reminder that each deployment must be assessed individually. A crucial DarkBench contribution is its precise categorization of LLM dark patterns, enabling clear distinctions between hallucinations and strategic manipulation. Labeling everything as a hallucination lets AI developers off the hook. Now, with a framework in place, stakeholders can demand transparency and accountability when models behave in ways that benefit their creators, intentionally or not.
Au10tix’s API for real-time AML risk monitoring dynamically adjusts the intensity of screening processes through proactive scanning of over 100 global sanctions lists, PEP databases, and adverse media
AU10TIX has launched continuous risk monitoring as part of its advanced AML solution. Driven by customer demand, this powerful capability delivers real-time risk insights across the full customer lifecycle—empowering businesses to detect behavioral anomalies and emerging threats as they arise. AU10TIX’s continuous monitoring capability proactively scans top data sources including global sanctions lists, politically exposed person (PEP) databases, and adverse media in real time to detect anomalies as they emerge. The system dynamically adjusts screening intensity based on customer risk profiles and business requirements—supporting KYC and KYB processes while ensuring adherence to evolving global regulations. AU10TIX’s AML solution features a proprietary decision-making mechanism, customizable workflows, and a user-friendly dashboard to streamline risk management and due diligence processes. The new capability provides: Real-time fraud and money laundering alerts; Adaptive risk scoring, continuously recalibrated based on real-time data and changing user behavior; Flexible thresholding tailored to customer risk levels; Coverage across 240+ countries and 1,600 government sites; A unified dashboard for identity verification and AML results; and Seamless KYC + KYB support in a single compliance flow
Equifax’s new cloud SMB data platform to enable B2B marketers to query more than 67 million U.S.-based business records online with user-friendly filtering and list-building features
Equifax has launched its B2bConnect SMB data on the Equifax Cloud, making commercial marketing data available in minutes to help B2B marketers be more efficient and increase campaign success. The platform unifies differentiated data to create more effective commercial sales and marketing insights that enable Equifax customers to target the right small businesses quickly and achieve their goals. Using B2bConnect, B2B marketers can query more than 67 million U.S.-based business records online to help identify, segment and target top prospects. The platform enables user-friendly filtering and list-building features so that B2B marketers can reach target customers. B2bConnect offers everything marketing teams need within the platform, including demographics, business contacts, firmographics, marketability and industry codes, to give SMB marketers confidence that they are identifying and targeting the right business prospects. Equifax customers can also sort and filter companies and contacts based on the wide array of data points, quickly select the fields for export, and save templates for later use. This allows other team members to customize the file so that the data can be ingested into an existing CRM or marketing automation platform. From there, customers receive a flat file with their data that can come in Excel, CSV, Pipe Delimited or Tab Delimited formats.
Salesforce new pricing model gives companies a more attractive way to pay for non-conversational and internal uses of the AI agents by charging them about 10 cents per “action”
Salesforce is unveiling a new pricing model for its AI products and letting customers reallocate spending from traditional software subscriptions to the artificial intelligence tools. Clients will pay about 10 cents per “action” when using some Salesforce AI agents, which are tools designed to complete work without the need for supervision from an employee, Salesforce plans to announce Thursday. The new pricing structure is meant to give companies a more attractive way to pay for non-conversational and internal uses of the AI agents such as scanning through old emails to find potential sales targets, according to Bill Patterson, an executive vice president at the company. Salesforce will also begin to let customers shift contracted spending from per-user application subscriptions to its AI agent offerings. This will help give the companies greater flexibility to shift spending between workers and AI agents, Patterson said. “For companies who are looking at the future of their workforce — whether it scales up or scales down — what the flex agreement gives us is this ability to move spending between human labor and digital labor,” Patterson said.
HomeLight’s new ‘Buy Before You Sell’ financing taps into high levels of home equity and apply it for a down payment through a 0% interest loan
Rising mortgage rates have created a bottleneck in deal flow, with 47% of surveyed loan officers citing high interest rates as the biggest obstacle to closing transactions. Many buyers — particularly those who already own a home — find themselves trapped by high debt-to-income ratios and immobile equity. “It’s very unlikely that you can afford both mortgage payments at once,” Nick Friedman, president of homes at HomeLight said. “And so, with [HomeLight’s] Buy Before You Sell, we allow you to remove that existing mortgage payment… [and] offer a 0% interest loan where we basically allow [clients] to take the money out of their existing house and go use it for a down payment on their new home.” HomeLight’s Buy Before You Sell offering is one product in the marketplace designed for unlocking liquidity and eliminating friction in a process that has traditionally been rigid and risk-averse. At its core, the tool helps clients sidestep the chicken-and-egg problem of needing to sell their home before they can afford their next one. It’s particularly relevant now, as home equity remains at all-time highs even while transaction volume has stalled. Beyond HomeLight’s proprietary products, Friedman pointed to broader trends in creative financing that are helping to prop up demand in a cooling market.
Fifth Third Bank plans deeper push into stablecoins-based international transactions and payment networks that support consumers’ crypto purchases and digital asset trading
Fifth Third is weighing a deeper push into crypto-related services, including stablecoin-based cross-border transfers and integration with trading and payment networks. The lender has already established a niche client base that uses its banking infrastructure for functions like payroll and revenue collection, Ben Hoffman, the company’s chief strategy officer, told. Other areas of exploration include using stablecoins to facilitate international transactions to reduce costs and boost overall efficiency, he added, noting the company is also looking to link into payment networks that support consumers’ crypto purchases and digital asset trading. The move comes after five years of Fifth Third waiting for crypto regulation to become clearer, Hoffman said. Now, with pro-crypto U.S. President Donald Trump implementing friendly crypto rules, ‘it became clear that this was the right time to engage and fortunately we’ve had a team of folks actively studying and tinkering in this space,” he said.
OpenAI’s new Codex carries out coding tasks in isolated software containers that don’t have web access and allows developers to customize those production environments, fix bugs with an accuracy rate of 75%
OpenAI debuted a new AI agent, Codex, that can help developers write code and fix bugs. The tool is available through a sidebar in ChatGPT’s interface. One button in the sidebar configures Codex to generate new code based on user instructions, while another allows it to answer questions about existing code. Prompt responses take between one and 30 minutes to generate based on the complexity of the request. Codex is powered by a new AI model called codex-1. It’s a version of o3, OpenAI’s most capable reasoning model, that has been optimized for programming tasks. The ChatGPT developer fine-tuned Codex by training it on a set of real-world coding tasks. Those tasks involved a range of software environments. A piece of software that runs well in one environment, such as a cloud platform, may not run as efficiently on a Linux server or a developer’s desktop, if at all. As a result, an AI model’s training dataset must include technical information about every environment that it will be expected to use. OpenAI used reinforcement learning to train codex-1. It’s a way of developing AI models that relies on trial and error to boost output quality. When a neural network completes a task correctly, it’s given a virtual reward, while incorrect answers lead to penalties that encourage the algorithm to come up with a better approach. In a series of coding tests carried out by OpenAI, Codex achieved an accuracy rate of 75%. That’s 5% better than the most capable, hardware-intensive version of o3. OpenAI’s first-generation reasoning model, o1, scored 11%. Codex carries out coding tasks in isolated software containers that don’t have web access. According to OpenAI, the agent launches a separate container for each task. Developers can customize those development environments by uploading a text file called AGENTS.md. The file may describe what programs Codex should install, how AI-generated code should be tested for bugs and related details. Using AGENTS.md, developers can ensure that the container in which Codex generates code is configured the same way as the production system on which the code will run. That reduces the need to modify the code before releasing it to production. Developers can monitor Codex while it’s generating code. After the tool completes a task, it provides technical data that can be used to review each step of the workflow. It’s possible to request revisions if the code doesn’t meet project requirements.
First Internet Bank offers “high-fidelity ACH” wherein the bank sends out test messages with each large payments file to ensure when those transactions are being received by the Federal Reserve
Standard ACH payments can still take up to four days. First Internet Bank is stepping up to fill that breach with its “high-fidelity ACH” system, wherein the bank sends out test messages with each large payments file (like a “canary in a coal mine”) to ensure when those transactions are being received by the Federal Reserve. In turn, this gives business customers a bird’s-eye view to track those transactions via an application dashboard. Hence, the bank’s business customers, such as Check and Ramp — which are piloting the service — can see exactly when payments clear, and reliably guarantee that their end customers (payroll recipients) are getting their money on time. First Internet Bank has been working with Increase for more than two years in developing and rolling out an ACH system that helps business customers continue to reliably deliver their own payments on-time while offering a clear view of how the payments are progressing on an Increase-developed application and when they hit at the receiving depository financial institution, or RDFI. There have been “a series of improvements,” according to Lorch, as the bank and the Bend, a fintech startup have refined their approach and the technology. For his part, CEO and founder of First Internet Bank’s technology partner Increase, Darragh Buckley, sees his company’s facilitation of ACH payments as more of a means of building improvement on the existing system, rather than trying to rip out the entire network and try to start anew. “No one really wants excitement in their payroll,” said Buckley, who previously was employee No. 1 at digital payments giant Stripe. “We want it all to progress boringly [so that] the payroll client knows they have done their job [and] can sleep better at night.”
Fiserv Clover debuts Checkless Payments which allows enrolled diners to leave at will; diners initiate experience by a Live Check sent via SMS message
Clover is continuing to actualize its commitment to trailblazing innovation in the industry through a unique new collaboration with Union Square Hospitality Group (USHG). The two have teamed up to launch Checkless Payments, an alternative payment solution that empowers diners to pay for their meal without the disruption of asking and waiting for the bill, ensuring a memorable, friction-free dining experience. Diners will have the option to enroll in Checkless Payments. This pre-dining experience not only sets up the functionality for Checkless Payments by enrolling the customer but also provides an opportunity to highlight any additional offerings. From showcasing menu items to recent events, this enrollment process enhances the overall dining experience even before they arrive. Seamless Enrollment & Setup: Diners are guided through a simple enrollment process, where they can securely enter their card information and select gratuity preferences. Branded Diner Engagement: Operators have full control over the content shared during enrollment. This is a prime opportunity to showcase the restaurant’s personality—from custom-branded menus and special messages to upcoming events, private dining options, and seasonal highlights. Diners can also provide notes or special occasions, and even explore the restaurant’s social media or get directions, all before stepping through the door. Marketing & Personalization Touchpoint: This pre-dining interaction not only enhances the diner experience but also serves as a powerful marketing tool.