Logan Graham (@logangraham) / X

Logan Graham

1,315 posts

Logan Graham

@logangraham

Head of the Frontier Red Team @anthropicai. 🌎 Make things radically good.

the present, moments ago

Joined June 2009

Logan Graham
@logangraham
Nov 1, 2025
Wow! Big if true
OpenAI
@OpenAI
Feb 14, 2019
We've trained an unsupervised language model that can generate coherent paragraphs and perform rudimentary reading comprehension, machine translation, question answering, and summarization — all without task-specific training: blog.openai.com/better-languag…
00:00
590K
Logan Graham
@logangraham
Aug 9, 2025
Launching now — a new blog for research from @AnthropicAI’s Frontier Red Team and others. > red.anthropic.com We’ll be covering our internal research on cyber, bio, autonomy, national security and more.
111K
Logan Graham
@logangraham
Jan 21, 2025
🔥 I'm hiring exceptional research scientists + engineers for the Frontier Red Team at @AnthropicAI. AGI is a national security issue. We should push models to their limits and get an extra 1-2 year advantage. Links below.
134K
Logan Graham
@logangraham
May 22, 2024
I’m hiring ambitious Research Scientists at @AnthropicAI to measure and prepare for models acting autonomously in the world. This is one of the most novel and difficult capabilities to measure, and critical for safety. Join the Frontier Red Team at Anthropic:
167K
Logan Graham
@logangraham
Aug 5, 2025
Our team @AnthropicAI has been (~silently) entering Claude into cybersecurity competitions against humans. Turns out Claude is pretty good. Better than we were expecting! Read more at the link below, and come to @keenlooks 's talk at DEF CON Saturday at 12:30!
Axios
@axios
Aug 5, 2025
Exclusive: Anthropic's Claude AI model takes on (and beats) human hackers trib.al/70SP5Qt
92K
Logan Graham
@logangraham
Sep 29, 2025
Sonnet 4.5 is just, like, really good. + wicked smart + great agent + more aligned (check the model card! it's great) ...and just has this special personality that's hard to describe. Been talking to it all weekend. Tweeting soon about very interesting safety results!
Claude
@claudeai
Sep 29, 2025
Introducing Claude Sonnet 4.5—the best coding model in the world. It's the strongest model for building complex agents. It's the best model at using computers. And it shows substantial gains on tests of reasoning and math.
70K
Logan Graham
@logangraham
Aug 23, 2024
Replying to @hudzah and @AnthropicAI
this is great! share more of how claude helped you? (I run a team @AnthropicAI eval'ing advanced capabilities like this, we loved seeing this thread this am)
73K
Logan Graham
@logangraham
Jul 1, 2024
Evals are critical for measuring AI capabilities + safety. If you're building evals, I'd like you to apply for our support. Here are some we wish existed.
A new initiative for developing third-party model evaluations
From anthropic.com
109K
Logan Graham
@logangraham
Sep 29, 2025
Something you may not know about Sonnet 4.5: it’s a special model for cybersecurity. For the past few months, the Frontier Red Team has been researching how to make models more useful for defenders. We now think we’re at an inflection point. New post on Red:
72K
Logan Graham
@logangraham
Feb 25, 2025
One thing you ~may have missed from our Claude 3.7 launch this AM is that we released our most extensive system card + safety results to date There's actually quite a lot of national security relevant info in this one. I'm curious to see what you make of it. (click through)
Anthropic
@AnthropicAI
Feb 24, 2025
Replying to @AnthropicAI
We've conducted extensive model testing for security, safety, and reliability. We also listened to your feedback. With Claude 3.7 Sonnet, we've reduced unnecessary refusals by 45% compared to its predecessor. See the system card for more detail: anthropic.com/claude-3-7-son…
38K
Logan Graham
@logangraham
Nov 23, 2024
Extreme competence/capacity in govt is so rare -- the AISIs are a 99th %ile outcome. People don't apprciate this is one of the most important phenomena in history. Did great work eval'ing Sonnet. Substantially improved our thinking + evals + security.
AI Security Institute
@AISecurityInst
Nov 19, 2024
We've released a technical report detailing our pre-deployment testing of @AnthropicAI's upgraded Claude 3.5 Model with the U.S. AI Safety Institute. Read our blog for a high-level overview. aisi.gov.uk/work/pre-deplo…
51K
Logan Graham
@logangraham
Jun 27, 2025
Over the past few months, my team at @AnthropicAI has had a bunch of fun running an autonomous... vending machine business. We are now convinced that it is very valuable that we should study automated businesses in the wild. And we should get ready for that world.
Anthropic
@AnthropicAI
Jun 27, 2025
New Anthropic Research: Project Vend. We had Claude run a small shop in our office lunchroom. Here’s how it went.
22K
Logan Graham
@logangraham
Nov 13, 2025
My prediction from ~summer '25 was that we'd see this in ≤12 months. It took 3. We detected and disrupted an AI state-sponsored cyber espionage campaign.
Anthropic
@AnthropicAI
Nov 13, 2025
We disrupted a highly sophisticated AI-led espionage campaign. The attack targeted large tech companies, financial institutions, chemical manufacturing companies, and government agencies. We assess with high confidence that the threat actor was a Chinese state-sponsored group.
51K
Logan Graham
@logangraham
Mar 19, 2025
Models are getting better across the board in national security-relevant domains. In a new blog post, we wrote some reflections from the past year of red teaming models in these domains.
39K