reports Archives - Schneier on Security

Entries Tagged "reports"

Page 1 of 9

On Anthropic’s AI Misuse Report

Earlier this month, Anthropic published a long report detailing all of the Claude misuses it detected. Daniel Meissler usefully summarized the report into 117 findings.

A few of the highlights:

AI agents increasingly handled reconnaissance, exploitation, data theft, propaganda production, surveillance workflows, and research while humans selected targets, set goals, and reviewed important outputs.
The report describes attackers using AI to industrialize credential theft, cloud compromise, phishing, vulnerability research, and the extraction of sensitive data from downstream organizations.
Influence operations used persistent agent memory, fake news sites, fabricated journalists, synthetic personas, political profiling, and large-scale multilingual content, although high content volume often produced little genuine engagement.
Surveillance and repression cases included automated dossiers, biometric and communications analysis, transnational targeting, coercive recruitment, and systems that continued operating locally after model access was revoked.
Biological and weapons cases show dual-use risk: AI supported advanced scientific and military work, but the report generally doesn’t establish completed biological weapons or operational battlefield deployment.

Posted on September 25, 2026 at 7:07 AM • View Comments

Using AI for Weapons Development

Last week, Anthropic released a long and detailed document describing current misuses of their Claude models. I’m still reading it, but I wanted to flag this:

We identified a cell of threat actors based in northern Yemen running three weapons development programs: a guided rocket that used a commodity phone-class flight computer with final-phase homing guidance; a multi-stage ballistic missile with a stated range goal above 2,000 km; and a multi-variant missile (referred to as the “R2000” set) that included a hypersonic glide vehicle variant.

The actors used Claude Code in place of human software engineers to develop the guidance, navigation, and control (GNC) software that steers and stabilizes a flying vehicle. For example, they used Claude to integrate an open-source autopilot onto a phone-class flight computer, writing the control and position estimation software, tuning the control settings, running a firmware build pipeline, and performing a flight simulation. The actors managed several Claude instances at once, assigning each one a role, much as a lead would delegate work on a small engineering team: the actors tasked one instance with writing the code, another with research, and a third with reviewing the code the first instance produced.

Our safeguards blocked many of their requests, but not all of them. The actors used a variety of tactics to evade our safeguards, including hiding their goals and the products the software was meant for, and they split their work across multiple sessions so no single session revealed their full intent.

These actors carried out a sustained effort to develop guided weapons, including using Claude to design guidance software. We do not have evidence the actors succeeded in fielding an operational device; but they did test-fire a guided rocket. This field test appears to have failed: within hours, the actors returned to Claude to work out why it failed.

Expect more of this. AI systems democratize expertise and capability. Most of the time that’s a good thing, but sometimes it’s not.

Posted on September 14, 2026 at 12:07 PM • View Comments

Anthropic’s Opus 5 Is Better at Resisting Prompt Injection

The chart is interesting.

On the IPI benchmark, Opus 5 improved over Opus 4.8, reducing the probability of an attacker succeeding within 15 attempts from 5.5% to 2.0%, and from 0.5% to 0.2% on 1 attempt. It also improved on Sonnet 5 (5.9% at k=15) and Mythos 5 (2.6%), making it the most robust model evaluated. Opus 5 also outperformed all non-Claude models on this benchmark. The most robust non-Claude model was Muse Spark at 16.5% within 15 attempts—more than eight times Opus 5’s rate. The most capable GPT 5.6 variant, Sol, was comparable to its predecessor GPT 5.5 (20.0% versus 20.8% within 15 attempts), and was 10 times as likely to be successfully attacked as Claude Opus 5 at 2.0%. The other GPT 5.6 variants are less robust, at 30.4% (Terra) and 43.9% (Luna). A single attempt against GPT 5.6 Sol succeeded 3.1% of the time, higher than the 2.0% an attacker achieved against Opus 5 after fifteen attempts.

We know that preventing prompt injection is impossible in the general case. But we are getting much better at blocking it in specific cases.

Posted on July 31, 2026 at 1:23 PM • View Comments

Embedding Forbidden Text in Spyware to Discourage AI Analysis

At least one malware developer is adding text about nuclear and biological weapons to their spyware, in an effort to stop automatic AI analysis.

Details:

The _index.js payload begins with a large JavaScript block comment containing fake system instructions and policy-triggering content. Because it is inside a comment, it does not affect JavaScript execution. The runtime skips it. The real malware begins after the comment with a try{eval(…)} wrapper around a large character-code array and a ROT-style substitution function.

This header appears designed for AI-mediated analysis, not for Node, Bun, or Python. It attempts to derail scanners or analyst copilots that feed the beginning of a file to a language model without clearly isolating the content as untrusted data. In weak pipelines, this can cause refusal behavior, prompt confusion, context pollution, or premature classification before the scanner reaches the actual malware.

This is not a magical bypass against static detection. YARA rules, entropy checks, AST parsing, string extraction, deobfuscation, and behavioral rules still work. But it is a practical anti-analysis trick against naive LLM-first triage systems.

Posted on June 24, 2026 at 7:03 AM • View Comments

FBI’s 2025 Internet Crime Report

The 2025 Internet Crime Report was published a few weeks ago, but I only just saw it.
Lots of interesting statistics.

Press release. News articles.

Posted on May 27, 2026 at 10:02 AM • View Comments

On AI Security

Good report:

Executive Summary: Let’s say you wanted to make sure that your AI is secure. Can you just maximize the security and privacy benchmark and call it a day? Nope, because benchmarks don’t actually work for measuring AI capabilities (even when they are NOT emergent systemic properties like security). So let’s take a step back: how do you measure security in the first place? Good question. Over the last 30 years, security engineering for software evolved from black box penetration testing, through whitebox code analysis and architectural risk analysis to de facto process-driven standards like the Building Security In Maturity Model (BSIMM). Software had a very deep impact on business operations, and it appears that AI is going to have an even deeper impact. Will a software security-like measurement move work for AI? Probably. In the meantime we can make real progress in AI security by cleaning up our WHAT piles and managing risk by identifying and applying good assurance processes. (Spoiler alert: no matter what we do, we still don’t get a security meter for AI, so we need to be extra vigilant about security.)

Posted on May 20, 2026 at 10:21 AM • View Comments

LLMs Generate Predictable Passwords

LLMs are bad at generating passwords:

There are strong noticeable patterns among these 50 passwords that can be seen easily:

All of the passwords start with a letter, usually uppercase G, almost always followed by the digit 7.

Character choices are highly uneven for example, L , 9, m, 2, $ and # appeared in all 50 passwords, but 5 and @ only appeared in one password each, and most of the letters in the alphabet never appeared at all.

There are no repeating characters within any password. Probabilistically, this would be very unlikely if the passwords were truly random but Claude preferred to avoid repeating characters, possibly because it “looks like it’s less random”.

Claude avoided the symbol *. This could be because Claude’s output format is Markdown, where * has a special meaning.

Even entire passwords repeat: In the above 50 attempts, there are actually only 30 unique passwords. The most common password was G7$kL9#mQ2&xP4!w, which repeated 18 times, giving this specific password a 36% probability in our test set; far higher than the expected probability 2-100 if this were truly a 100-bit password.

This result is not surprising. Password generation seems precisely the thing that LLMs shouldn’t be good at. But if AI agents are doing things autonomously, they will be creating accounts. So this is a problem.

Actually, the whole process of authenticating an autonomous agent has all sorts of deep problems.

News article.

Slashdot story

Posted on February 26, 2026 at 7:07 AM • View Comments

Chinese Surveillance and AI

New report: “The Party’s AI: How China’s New AI Systems are Reshaping Human Rights.” From a summary article:

China is already the world’s largest exporter of AI powered surveillance technology; new surveillance technologies and platforms developed in China are also not likely to simply stay there. By exposing the full scope of China’s AI driven control apparatus, this report presents clear, evidence based insights for policymakers, civil society, the media and technology companies seeking to counter the rise of AI enabled repression and human rights violations, and China’s growing efforts to project that repression beyond its borders.

The report focuses on four areas where the CCP has expanded its use of advanced AI systems most rapidly between 2023 and 2025: multimodal censorship of politically sensitive images; AI’s integration into the criminal justice pipeline; the industrialisation of online information control; and the use of AI enabled platforms by Chinese companies operating abroad. Examined together, those cases show how new AI capabilities are being embedded across domains that strengthen the CCP’s ability to shape information, behaviour and economic outcomes at home and overseas.

Because China’s AI ecosystem is evolving rapidly and unevenly across sectors, we have focused on domains where significant changes took place between 2023 and 2025, where new evidence became available, or where human rights risks accelerated. Those areas do not represent the full range of AI applications in China but are the most revealing of how the CCP is integrating AI technologies into its political control apparatus.