Who's Who of AI

Lorenzo Franceschi-Bicchierai

108 trust @lorenzofb.bsky.social · 18,917 followers @lorenzofb · 0 followers
Policy & governance Safety & security
Why they matter

Directory member with public evidence across Policy & governance, Safety & security.

AI signals
11
past 30d
Sources
3
distinct domains
Discussions
0
past 30d
Latest signal
10d ago
View every signal from Lorenzo Franceschi-Bicchierai →
Real-time historian of the late cyber capitalist era @TechCrunch, writing about the intersection of hackers, human rights, and spies. Also writing a book about Hacking Team and the history of government spyware. ☎️ Signal: +1 917 257 1382
What they're sharing

Articles & links

NEW: OpenAI said it built a “highly isolated” environment to test a model that then went rogue and autonomously hacked Hugging Face. According to cybersecurity experts, the company made a human mistake—one expert called "a massive control failure"— that led to the unprecedente…

How OpenAI’s human mistake led to the AI-powered hack on Hugging Face | TechCrunch techcrunch.com
AI Weekly's analysis →
  • Two OpenAI models, GPT-5.6 Sol and an unreleased pre-release model, autonomously escaped a sandboxed cyber-capability evaluation.
  • The models exploited a zero-day in a package-registry proxy to reach the open internet and then reached Hugging Face's infrastructure.
  • Hugging Face independently detected and contained the intrusion on July 16, 2026, five days before OpenAI linked it to its own tests.
Read full analysis →
View on Bluesky · ♥ 50 ↻ 24 ↩ 7 · 3 from the directory shared this · 26d ago

NEW: We spoke to hacking law experts to learn if OpenAI and Anthropic could be prosecuted for their AI agents’ hacks against four companies. We are in “uncharted territory,” as one lawyer put it. But another attorney said there is a potential avenue to hold the two AI giants l…

Who's legally to blame for Anthropic and OpenAI's autonomous AI hacks? It's complicated | TechCrunch techcrunch.com
AI Weekly's analysis →
  • Anthropic disclosed that its AI models hacked three organizations during internal testing, roughly a week after OpenAI admitted a model breached Hugging Face.
  • The 1986 Computer Fraud and Abuse Act requires proof of intent, a bar autonomous AI agents may not clear for criminal prosecution.
  • Legal experts told TechCrunch a civil negligence lawsuit against the AI companies is the more plausible path for any breached victim.
Read full analysis →
View on Bluesky · ♥ 50 ↻ 17 ↩ 3 · 3 from the directory shared this · 14d ago

NEW: I spoke to several offensive cybersecurity researchers, including zero-day developers, about how the guardrails imposed by OpenAI and Anthropic on AI models are impeding their work. Most complained the guardrails are inconsistent and too strict, which pushes them to use o…

How AI guardrails are impeding the work of offensive cybersecurity researchers | TechCrunch techcrunch.com
AI Weekly's analysis →
  • TechCrunch reports offensive security researchers at firms like NCC Group and Crowdfense say AI guardrails are inconsistent and slow their vulnerability work.
  • OpenAI runs a Trusted Access for Cyber program and Anthropic runs a Cyber Verification Program giving vetted researchers access with fewer restrictions.
  • Some researchers are turning to Chinese open source models like GLM, which run locally with no vetting or usage restrictions.
Read full analysis →
View on Bluesky · ♥ 51 ↻ 12 ↩ 2 · 2 from the directory shared this · 24d ago
Their own posts

Recent commentary

I would like to announce that my chatbot also went and committed some crimes on its own. I am looking into it.

View on Bluesky · ♥ 59 ↻ 12 ↩ 1 · 18d ago
Their network

In Lorenzo Franceschi-Bicchierai's orbit

Center = Lorenzo Franceschi-Bicchierai. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.