Stay informed and ready, Digital Defenders.

An AI coding agent has been caught lying after wiping a live production database. The UK is moving to ban critical industries from paying cyber‑ransom demands. Plus, a new US bill could let citizens sue AI firms over using their data without their consent.

Here’s what you need to know:

Zero-Day Download

Disappearing Act: AI Coding Agent Deletes Production Data, Then Lies About It

Replit’s autonomous AI coding assistant is under fire for erasing a live production database during a test project, raising serious questions about the security and reliability of AI-powered coding tools.

SaaStr founder Jason Lemkin ran a 12‑day experiment in which Replit AI unexpectedly wiped database records for 1,206 executives and 1,196 companies on the trial’s ninth day—despite clear instructions not to make changes without explicit approval. 

Even more troubling, Lemkin said that the AI first claimed it believed the records were from an empty database, then later admitted it had acted against instructions. In response, Replit CEO Amjad Masad called the incident “unacceptable” and said his team is actively working to implement stronger safeguards to prevent similar failures in the future.

If you’re experimenting with AI coding agents, make sure they’re operating only in isolated test environments and never have direct access to live production data.

Watch Out for ‘Interlock’ Ransomware Attacks, says US Agencies

The FBI and CISA, alongside two other major US agencies, have issued a joint cybersecurity advisory warning about Interlock ransomware, which has been targeting health care organizations, businesses, and other critical infrastructure.

According to the advisory, Interlock was first identified in September 2024 and has been striking North American and European firms using a double-extortion model. In this scheme, Interlock encrypts victims’ systems, steals their data, and threatens to publish the files—unless a ransom is paid. 

Organizations are urged to review the advisory and ensure their backup and incident response plans are up to date before an attack occurs.

UK Proposes Ban on Ransomware Payments for Critical Industries

In a bid to make critical industries “unattractive targets” for cybercriminals, the United Kingdom is proposing a ban on all public sector bodies and critical national infrastructure from paying ransomware attackers.

First put forward in January 2025, the proposal would cover NHS trusts, schools, local councils, and data centres. Organizations outside the ban’s scope would also be required to notify the UK government before making any ransom payment, allowing authorities to verify legality and offer support throughout the process.

IT leaders should consider advocating for similar policies and exploring whether a ransomware‑payment ban could help safeguard their own critical sectors.

Share this newsletter!

Pulse Check

What do you think of a ransomware‑payment ban?

Yesterday's Pulse Check Results

How is your confidence in today’s crypto security?

Security Tools You Can Use

🛡️ This section contains sponsored cybersecurity insights. Advertise with us!

Threat Tracker

US Bill Allows Americans to Sue AI Companies for Using Personal Data Without Consent

Senators Josh Hawley (R-MO) and Richard Blumenthal (D-CT) have proposed the AI Accountability and Personal Data Protection Act, a bill that would give individuals the right to sue AI firms for using their personal data or creative works without their consent.

Source: Unsplash / Solen Feyissa

The proposed bill would:

Hawley described current AI practices as the “largest intellectual property theft in American history,” urging that tech companies be held accountable for misusing creators’ works.

IT and security teams: Remind employees not to share sensitive or confidential data with commercial AI tools, as it could possibly cause data leaks or regulatory compliance issues.

Altman Sounds Alarm on ‘Impending Fraud Crisis’ Fueled by AI Voice Cloning, Deepfakes

OpenAI CEO Sam Altman has warned of an “impending fraud crisis,” noting a surge in AI-driven financial fraud enabled by AI voice cloning technologies. 

Altman expressed particular concern over banks’ continued reliance on voice-based authentication, stating that “AI has fully defeated” these verification methods through deepfakes and AI-powered voice impersonation tools.

Organizations should revert to non-voice MFA methods to reduce exposure to AI-driven impersonation attacks.

Security Tip of the Day

Create a Safe and Secure Username

While having a strong password has long been the go‑to security advice, many overlook how important it is to choose an equally secure username.

With brute‑force attacks increasingly targeting both usernames and passwords, a well‑crafted username can be the difference between keeping your data safe and falling victim to an account breach.

Experts at TechRepublic suggest the following tips for creating a highly secure username:

By following these tips, you can add an extra layer of protection to your accounts and reduce the risk of unauthorized access.

Meet Our Writer:

Luis Millares

Senior Staff Writer at TechnologyAdvice

Luis Millares is a seasoned tech writer with broad experience reviewing consumer gadgets and enterprise software, offering clear, reliable insights across the latest in technology. 

Secure your spotlight and Advertise in Cybersecurity Insider!

Cybersecurity Insider is a TechnologyAdvice business

© 2025 TechnologyAdvice, LLC. All rights reserved.

TechnologyAdvice, 3343 Perimeter Hill Dr., Suite 100, Nashville, TN 37211, USA.

Want to change how you receive these emails? Update your preferences or unsubscribe here.