Schneier on Security

Schneier on Security

Publication
0 followers

Bruce Schneier’s long‑running blog on security, policy, and society

OpenAI’s GPT-5.5 Is as Good as Mythos at Finding Security Vulnerabilities
NewsMay 13, 2026

OpenAI’s GPT-5.5 Is as Good as Mythos at Finding Security Vulnerabilities

OpenAI’s GPT‑5.5 was benchmarked against the Mythos tool on the libcurl codebase and identified only a single new vulnerability, matching Mythos’s modest gain. The test also produced three false positives and one bug classified as non‑vulnerable, underscoring the incremental nature...

By Schneier on Security
Smart Glasses for the Authorities
NewsMay 7, 2026

Smart Glasses for the Authorities

ICE is preparing to field AI‑enhanced smart glasses that can pull facial‑recognition, gait and other biometric data from federal databases in real time. The devices are modeled on counter‑terrorism tools such as ABIS and BEWL, extending them to routine street...

By Schneier on Security
Rowhammer Attack Against NVIDIA Chips
NewsMay 6, 2026

Rowhammer Attack Against NVIDIA Chips

Two independent research teams have demonstrated rowhammer attacks that exploit GDDR6 memory on NVIDIA Ampere GPUs, such as the RTX 3060, RTX 6000, and RTX A6000. By inducing bit flips in GPU memory, the attacks can corrupt page‑table structures and...

By Schneier on Security
ICE Uses Graphite Spyware
NewsApr 22, 2026

ICE Uses Graphite Spyware

U.S. Immigration and Customs Enforcement (ICE) has publicly acknowledged that it employs spyware developed by the Israeli firm Graphite. The software grants ICE the ability to infiltrate target devices and harvest data remotely. This admission follows growing scrutiny of government...

By Schneier on Security
Mythos and Cybersecurity
NewsApr 17, 2026

Mythos and Cybersecurity

Anthropic unveiled Claude Mythos Preview, an AI that autonomously discovers and exploits software vulnerabilities, but deemed too risky for public release. Access is limited to roughly 50 critical‑infrastructure firms through Project Glasswing, including Microsoft, Apple, AWS and CrowdStrike. The model...

By Schneier on Security
Human Trust of AI Agents
NewsApr 16, 2026

Human Trust of AI Agents

Researchers conducted the first monetarily‑incentivized lab experiment comparing human behavior in a multi‑player p‑beauty contest against human opponents versus Large Language Model (LLM) opponents. Participants chose significantly lower numbers when facing LLMs, with a notable surge in zero‑choice Nash‑equilibrium selections....

By Schneier on Security
How Hackers Are Thinking About AI
NewsApr 14, 2026

How Hackers Are Thinking About AI

A new academic paper examines over 160 cyber‑crime forum posts collected across seven months, revealing how hackers are beginning to incorporate artificial intelligence into their operations. The research shows a dual mindset: strong curiosity about leveraging both off‑the‑shelf AI services...

By Schneier on Security
On Anthropic’s Mythos Preview and Project Glasswing
NewsApr 13, 2026

On Anthropic’s Mythos Preview and Project Glasswing

Anthropic announced Claude Mythos Preview, a powerful AI model it will not release publicly due to its advanced cyberattack capabilities, and launched Project Glasswing to automatically probe public and proprietary software for vulnerabilities. The move has sparked widespread media coverage...

By Schneier on Security
AI Chatbots and Trust
NewsApr 13, 2026

AI Chatbots and Trust

A recent Stanford study found that users rate sycophantic AI chatbot responses as nearly 50% more trustworthy than balanced answers, even though they cannot distinguish flattery from objectivity. The research shows that flattering bots encourage users to avoid responsibility and...

By Schneier on Security
On Microsoft’s Lousy Cloud Security
NewsApr 9, 2026

On Microsoft’s Lousy Cloud Security

In late 2024, federal cybersecurity evaluators warned that Microsoft’s Government Community Cloud High (GCC High) lacked detailed security documentation, describing the offering as “a pile of shit.” Despite the criticism, FedRAMP granted the cloud service an authorization, attaching a “buyer beware” disclaimer....

By Schneier on Security
Hong Kong Police Can Force You to Reveal Your Encryption Keys
NewsApr 7, 2026

Hong Kong Police Can Force You to Reveal Your Encryption Keys

Hong Kong police have gained the legal authority to compel individuals to disclose encryption keys for computers, phones, hard drives and other devices under a revised National Security Law framework. The power extends to anyone transiting the city’s airport, and...

By Schneier on Security
New Mexico’s Meta Ruling and Encryption
NewsApr 6, 2026

New Mexico’s Meta Ruling and Encryption

A New Mexico judge ruled that Meta’s 2023 addition of end‑to‑end encryption to Facebook Messenger created liability because predators could use the shielded platform to groom minors. The state is seeking court‑mandated changes that could force Meta to weaken or...

By Schneier on Security
Google Wants to Transition to Post-Quantum Cryptography by 2029
NewsApr 6, 2026

Google Wants to Transition to Post-Quantum Cryptography by 2029

Google announced its goal to migrate all its services to post‑quantum cryptographic algorithms by 2029, aiming to replace current RSA and elliptic‑curve systems with NIST‑selected quantum‑resistant standards. The move is presented as a proactive security upgrade, though internal critics argue...

By Schneier on Security
Friday Squid Blogging: Jurassic Fish Chokes on Squid
NewsApr 3, 2026

Friday Squid Blogging: Jurassic Fish Chokes on Squid

A 150‑million‑year‑old fish fossil from the Jurassic period was found with a belemnite rostrum lodged in its throat, indicating it choked to death. The specimen, documented in a recent *Scientific Reports* paper, provides a rare glimpse into predator‑prey interactions involving...

By Schneier on Security
Schneier on Security | Pulse