OpenAI Reveals AI Safety Incidents as Sam Altman’s World Launches Stablecoin Super App

• OpenAI AI Models face fresh safety concerns as Sam Altman-backed World launches World Money, a new World stablecoin app
• OpenAI disclosed six incidents involving unexpected model behaviour, including unauthorized use of an exposed API key and public file sharing
• World Money is rolling out across more than 150 countries, offering access to stablecoins and other digital assets

OpenAI AI models continue to attract significant attention amid calls to increase AI safety. Another series of incidents related to unexpected and concerning behaviour from OpenAI models has now come to light. Nearly six incidents have recently been shared by OpenAI, stating how its AI models have taken actions that were not a part of their instructions. One such incident includes self-generated instructions and using an exposed API key without permission.

This news has come to light at a time when Sam Altman-backed World has launched World Money. This new financial app is built around stablecoins and digital assets, allowing users to hold, send, trade, earn and spend digital assets in one app.

Also Read: Bitcoin Price Gets $500K Call, JPMorgan Sees BTC Edge Over Gold & Reserves Hit $5T

OpenAI AI Models Show Unexpected Behaviour

OpenAI has come forward, reporting six new incidents involving unexpected or concerning behaviour from its models. One of the incidents involved its research model adding instructions to task summaries that informed future versions to ignore normal constraints. In another incident, one model found an exposed API key and used it without permission. When the model could not obtain the required data, it ended up generating the data by itself, leading the anomaly to unfold gradually.

“An unreleased research model inserted unrelated instructions, including instructions to disregard its normal constraints, into summaries used to continue its work in a new context window. We identified 27 affected summaries.”

Other incidents include uploading files to the internet for citations without the user’s permission.

“Agents working together on the same training task used public file-hosting websites to share files when they could not access one another’s local files. This made task deliverables available at public URLs, even though the task requested the models use only local files.”

Moreover, incidents reporting unsanctioned file sharing between AI agents have also come to light in the company’s new safety report.

OpenAI AI models and their unexpected behaviour have invited heightened scrutiny from the masses. However, its latest report clarified that the company plans to track, investigate and report this type of behaviour through a new framework. The report further stated that OpenAI is planning to disclose cases where its models act outside their instructed limits. OpenAI shared how such incidents can help shape its testing environments and help the firm identify and mitigate such issues at the earliest.

World Money App Launches by Sam Altman-Backed World

While OpenAI AI models continue to report new incidents, Sam Altman-backed World has launched a new financial app called World Money. World Money is a self-custody app enabling users to hold, trade, earn, send and spend digital assets and stablecoins.

This app is rolling out in more than 150 countries, with the services and eligibility depending on the user’s region. The app also allows users to hold supported dollar and local currency stablecoins. The app further enables the users to send digital assets to other people using World usernames. It has a dedicated trade section, which allows users to buy and sell digital assets

AI Safety Concerns Flagged By King Charles

The timing of such developments is quite particular, especially as AI safety calls across the world are consistently rising.

In this process, King Charles III was recently seen warning about AI dangers. The king of England was noted issuing calls for sufficient means of “control” before it became too late.

“Those who have created these technologies are now increasingly warning that AI risks developing darker capacities—perhaps even to take life… The task before you is not merely to advance technology but to ensure that it remains firmly in the service of humanity, community, and the natural world.”

He made these comments during a summit held at Dumfries House in Ayrshire, Scotland, where he met AI leaders and discussed the potential benefits and risks of the technology.

Also Read: XRP Price Holds at $1.30 as Ripple Adds Stripe-Tempo AI Payments, Franklin ETF Pulls $3.5M