BTC $71,807
2026 Bull Run Is Building Start trading with 5% OFF all fees
Sign Up Now
BTC $71,807
Bull Run 2026 | 5% Off Fees Open your Binance account today
Sign Up

Microsoft Unveils Scanner to Detect Backdoored AI Models

Microsoft scanner detects hidden backdoors in open AI models without prior knowledge or training.

  • Microsoft has created a new scanner designed to find hidden backdoors in open-weight Large Language Models (LLMs).
  • The tool relies on three distinct behavioral signals within a poisoned model, such as unusual “double triangle” attention patterns.
  • This AI security breakthrough operates without needing prior knowledge of the backdoor or additional model training.
  • The scanner is a practical move to improve trust in AI, as Microsoft simultaneously expands its Secure Development Lifecycle to address AI-specific threats.

On February 4, 2026, Microsoft‘s AI Security team announced it built a lightweight scanner to detect backdoors in open-weight AI models. This development aims to improve overall trust in artificial intelligence systems by identifying covertly poisoned “sleeper agent” models.
The scanner leverages three observable signals to reliably flag backdoors with a low false positive rate, researchers said. These signatures are grounded in how trigger inputs measurably affect a model’s internal behavior, providing a technically robust basis for detection.
One key signal is a distinctive “double triangle” attention pattern that causes the model to focus on a trigger. Another is that backdoored models tend to leak their own poisoning data via memorization rather than training data.
Consequently, the scanner first extracts memorized content and then analyzes it to isolate suspicious substrings. Finally, it formalizes the three signatures as loss functions, scoring and ranking potential triggers.
However, the scanner does not work on proprietary models as it requires access to the model files. It also works best on trigger-based backdoors that generate deterministic outputs and is not a cure-all solution.
Meanwhile, Microsoft said it’s expanding its Secure Development Lifecycle to address AI-specific security concerns. “Unlike traditional systems with predictable pathways, AI systems create multiple entry points for unsafe inputs,” corporate VP Yonatan Zunger said. These entry points can carry malicious content or trigger entirely unexpected behaviors.

- Advertisement -

✅ Follow BITNEWSBOT on Telegram, Facebook, LinkedIn, X.com, and Google News for instant updates.

Previous Articles:

- Advertisement -
Ad
Altseason Is Loading. Don't watch from the sidelines.
SOL $90.51
DOGE $0.0963
LINK $9.02
SUI $1.00
5% off fees when you sign up
Start Trading
Ad
Pay Less on Every Trade. For Life.
$10K/mo volume Save $60/yr
$50K/mo volume Save $300/yr
$100K/mo volume Save $600/yr
5% off all trading fees when you sign up
Claim Your Discount

Latest News

ImmunityBio founder touts sepsis trials, NantX after Binance theft

Patrick Soon-Shiong renewed his “NantX coming soon” teaser after a $6.3 billion Binance theft,...

Crypto.com Shuts Hong Kong App, Eyes Institutional Exchange

crypto.com will shut down its consumer app in Hong Kong on Feb. 1, 2027,...

Kim Yo Jong denies North Korea behind Binance hack, hints US

Kim Yo Jong denied North Korea's involvement in the $6.3 billion Binance hack and...

Strive buys 124 BTC, fourth among public companies

Strive acquired 124 Bitcoin for roughly $10 million between Oct. 24 and Oct. 30,...

Claude finds 9.8 zero-click Linux root bug

Claude discovered a zero-click Linux vulnerability (CVE-2026-98432) rated 9.8 out of 10, described as...

Must Read

How to Swap Ethereum for Monero Without an Account

Paste your main Ethereum address into a block explorer and look at what a stranger can read off it. Every transfer in and out....
Ad
Altseason Is Loading. These 4 coins are trending right now.
SOL $92.12
DOGE $0.0950
LINK $9.02
SUI $1.02
5% off spot fees when you sign up
Start Trading