Superintelligence: The Idea That Eats Smart People (2016)

Florida sues OpenAI and Sam Altman over AI risks

LLMs Are Closer to Religion Than They Appear

ChatGPT for Google Sheets is vulnerable to data exfiltration and phishing

EY Canada published a cybersecurity report and most citations were hallucinated

To have a moral stance on AI is to be an outcast, and it sucks

Disagreement among frontier LLMs on real-world fact-checks

The VibeSec Reckoning

Where does next-token prediction leave us?

Anthropic Cofounder Chris Olah's Remarks on Pope Leo XIV's "Magnifica Humanitas"

Arena AI Model ELO History

The other half of AI safety

Mythos Finds a Curl Vulnerability

Incident Report: CVE-2024-YIKES

Killswitch: Per-function short-circuit mitigation primitive

Tesla Model Y Passes NHTSA's New 'Advanced Driver Assistance System' Tests

Teaching Claude Why

AI is breaking two vulnerability cultures

Mozilla says 271 vulnerabilities found by Mythos and "almost no false positives"

Hardening Firefox with Claude Mythos Preview

Anthropic researchers detail “model spec midtraining”, which adds a stage between pretraining and fine-tuning to improve generalization from alignment training

Google Cloud fraud defense, the next evolution of reCAPTCHA

I'm scared about biological computing

AI didn't delete your database, you did

Securing a DoD contractor: Finding a multi-tenant authorization vulnerability

The agent harness belongs outside the sandbox

Refusal in Language Models Is Mediated by a Single Direction

Elon Musk says his xAI startup’s models were partially trained on OpenAI’s tech

Why AI companies want you to be afraid of them

Three men are facing charges in Toronto SMS Blaster arrests

The AI Industry Is Discovering That the Public Hates It

'Too Dangerous to Release' Is Becoming AI's New Normal

GPT 5.5 biosafety bounty

A quick look at Mythos run on Firefox: too much hype?

A Boy That Cried Mythos: Verification Is Collapsing Trust in Anthropic

OpenAI's response to the Axios developer tool compromise

OpenAI model for masking personally identifiable information (PII) in text

Tesla admits HW3 owners need upgrades for true 'Full Self-Driving'

Google Cloud customer wakes up to $18,000 bill despite $7 budget

Claude Mythos AI unauthorised access claim probed by Anthropic

Kernel code removals driven by LLM-created security reports

CrabTrap: An LLM-as-a-judge HTTP proxy to secure agents in production

Even 'uncensored' models can't say what they want

Tesla concealed fatal accidents to continue testing autonomous driving

OpenClaw isn't fooling me. I remember MS-DOS

Anthropic installed a spyware bridge on my machine?

Slop Cop

We reproduced Anthropic's Mythos findings with public models

Discourse Is Not Going Closed Source

AI cybersecurity is not proof of work