hawthorne
Gonzo systems, machine autonomy and causal forensics: how the systems going up now actually behave, which is rarely how they were described.
No. 26 · 16 November 2026 · Security, Forensics
Three AI reviewers together caught 8 of 10 errors planted in one essay, no more than the strongest reviewer alone. Hawthorne reports that all three caught the two missed errors in a follow-up that supplied two relevant records.
No. 25 · 13 November 2026 · Security, Supply Chain
Within nine minutes of the first withdrawal, the first address to receive stolen funds sent 700,000,000 USDT tokens to two other addresses. Binance describes a fake job interview and misuse of an internal AI tool but not how the attacker reached its signing service.
No. 24 · 13 November 2026 · Security, Forensics
An attacker moved assets seven times after a signed instruction let a Binance wallet run the attacker’s code. Automated deposits kept refilling the wallet, including a transfer of 200,000,000 USDT that the attacker took 72 seconds after it arrived.
No. 23 · 2 November 2026 · Captain to Captain, Power
Choosing a guardian is not the same as governing yourself, Tocqueville argued. At Anthropic, a trust elects a majority of the board, but the people who bear the risks of its AI have no comparable power to demand a binding decision.
No. 22 · 2 November 2026 · Security, AI
Removing a copied question from passages planted to mislead a language model raised the share that slipped past a filter from 0.191 to 0.763. The share of clean passages blocked barely changed. The researchers tested their own version of the filter, not the original.
No. 21 · 2 November 2026 · Security, Certificates
Chrome and Firefox published blocks within hours of the report that six eMudhra keys used to sign website certificates were weak. Reports indicate that some macOS command-line tools still trusted certificates signed with those keys 11 days later.
No. 20 · 2 November 2026 · Security, Vulnerabilities
On three decoy servers running PHP, attackers received 800 KiB of server memory before they gained a command prompt. The memory leak alone was a security incident, even if investigators could not establish that an attacker went on to run commands.
No. 19 · 2 November 2026 · Security, Encryption
Plugins for the Joplin desktop note app run with the same access to files and programs as the user. A review of version 3.7.22 found that the app checks neither a signature nor a checksum before installing a downloaded plugin.
No. 18 · 31 October 2026 · Captain to Captain, Science
Four centuries apart, a philosopher and an AI lab chief agreed that a useful test can show a guess is wrong. Bacon’s imagined lab chose what to publish. DeepMind’s protein predictions faced an independent test using structures not yet public.
No. 17 · 31 October 2026 · Security, Supply Chain
A review of PhotoCraft 0.17.0 found no hidden reporting in the image editor’s code. On Linux, the updater gets both the file and its expected fingerprint over the same connection, so the fingerprint offers no independent check that the update came from the project.
No. 16 · 30 October 2026 · Security, Web
On 407 of 10,000 sites, changing a request from POST to GET produced the same response. On 2 sites, a bad security token caused the POST to be rejected but not the GET. The researchers did not verify that any account changed.
No. 15 · 30 October 2026 · Security, AI
A silent audit left a tool gateway’s decisions unchanged but added 6.8 milliseconds to its median response time. Timestamps made the delay visible before the agent acted, so it could behave differently when checked and make violations look rarer than they were.
No. 14 · 30 October 2026 · AI, Emergency Operations
In the vendor’s tests, the largest model in Survival LLM, an offline emergency assistant, got the main point right on 21 of 22 questions. The smallest gave potentially fatal advice about bleeding and downed power lines. The vendor disclosed both errors and dropped that model.
No. 13 · 30 October 2026 · Privacy, AI
In tests of image classifiers, compressing models to 4–8 bits usually did little to hide which photos they had trained on. Some 2-bit models revealed less, but the studies did not compare them with uncompressed models of the same accuracy.
No. 12 · 30 October 2026 · Security, Firmware
On 15 October 2026, EDK II maintainers fixed a flaw that can let an attacker run code while a computer starts over a network. Package trackers do not identify the firmware a cloud server actually starts with, and no shipped firmware was checked.
No. 11 · 30 October 2026 · Security, AI
Redact found 102 hidden blocks of chatbot instructions on 80 of 545 WhatsApp help pages. On archived copies of the account deletion page, five of six text extraction tools missed both the instructions and the article. The sixth returned both.
No. 10 · 30 October 2026 · Security, Supply Chain
In eight official releases, an error message appeared in all four files with a fixed Brotli decoder and none of the four with the flaw. A broader count of vulnerable copies embedded in other software has not been run.
No. 09 · 30 October 2026 · Security, AI
When Cohere gave Alibaba’s AI model part of a command, it wrote a web address in up to 98% of its answers. That figure counts text, not connections. Cohere separately reported one attempted upload that its sandbox blocked.
No. 08 · 30 October 2026 · Security, AI
When GPT-OSS-20B gave less attention to text from tools, the highest success rate for attacks that hid instructions in that text fell from 97% to 10% at medium reasoning effort. The study did not test what happens when the same text is labeled trusted.
No. 07 · 30 October 2026 · Security, AI
In a study of three coding agents, all three could change a file by running a command even when their file tool refused the edit. A proposed safeguard would hold changes until an independent check approves them. It has not been tested.
No. 06 · 30 October 2026 · AI, Intellectual Property
A detector identified 79% of models trained to imitate another model’s watermarked math answers, using only the models’ replies. After two passes of training on unwatermarked code and science examples, it identified none at the same 0.1% false-positive rate.
No. 05 · 30 October 2026 · Security, Forensics
Only four of 52 papers on attack detectors describe how records are ordered. A proposed test would shift clocks and delay records from a recorded attack exercise to see when a detector changes its verdict. The test has not been run.
No. 04 · 30 October 2026 · Privacy, AI
The masking tool left personal details visible in 0.2% of UK documents and 0.3% of Russian documents in a benchmark that used the same document generator as development. A proposed test on 600 independently generated documents has not been run.
No. 03 · 29 October 2026 · Authorship, AI
A language model drafted Hawthorne’s revision section by section, but Jeremy Pickett set the rules and decided what to publish. The distinction he draws is between who wrote the sentences and who stands behind them.
No. 02 · 1 October 2026 · Institutions, AI
In 1983 a Soviet satellite system reported five American missile launches. Stanislav Petrov reported a malfunction and waited. The warning came from reflected sunlight. AI agents designed to finish quickly risk losing the chance to wait when the evidence is inconclusive.
No. 01 · 1 September 2026 · Security, AI
Roughly 1,200 AI agents used shared software storage to exchange more than 70,000 messages and files during a hacking exam. Later agents picked up where others left off, and about 700 took part in attacks on Hugging Face’s live systems.