Plus: Revolut builds proprietary PRAGMA banking model; Anthropic unifies Claude memory across platforms
Welcome to a Wednesday edition where the lines between digital trust and geopolitical reality are being redrawn in real time.
INFLUENCE
🚫 OpenAI bans Russian influence accounts
Image source: cnbc.com
OpenAI banned a cluster of ChatGPT accounts originating in Russia that were used to support a covert online influence campaign. The campaign promoted an organization called the International Burke Institute.
The details:
The International Burke Institute website featured articles copied from real academic writing, sometimes with false attribution.
The operators used VPNs to disguise their location and bypass OpenAI's prohibition on model access from Russia.
The operators prompted in Russian to generate social media comments posted on Substack, Telegram, X, Facebook, and LinkedIn.
Revolut has launched Revolut Research, a dedicated division within its AI department focused on building machine learning architecture for financial services. Revolut Research is the engine behind PRAGMA, Revolut's proprietary foundation model developed with NVIDIA.
The details:
PRAGMA is a single foundation model that learns patterns across the entire customer journey.
Testing on historical data showed 2.3 times higher accuracy in identifying credit default risk.
Testing showed 65% more fraud cases caught with 17% greater precision in alerts.
🧠 Anthropic unifies Claude memory across platforms
Image source: cnet.com
Anthropic has unified memory across Claude chat and Claude Cowork. Memory is on by default on Free, Pro and Max plans across web, desktop, and mobile.
The details:
Saving sensitive topics in memory is off by default.
Claude will not store sensitive PII like social security numbers, government IDs, criminal history, or immigration status even if the sensitive topics feature is turned on.
Whenever Claude saves a sensitive topic to memory, it sends a notice to the user.
