AI Security Institute (AISI) ✓
@AISecurityInst · United Kingdom · joined 14 Feb 2024
We conduct scientific research to understand AI’s most serious risks and develop and test mitigations.
20 635Followers
32Following
491Posts total
1.3MViews on collected posts
โพสต์ล่าสุด
กำลังค้นหาโพสต์ใหม่…
Some frontier AI evaluations require hundreds of millions of tokens, and every wasted trial has a real cost.
We’re introducing optstop: our open-source tool that stops an evaluation where estimates are already precise, and keeps running where they're not. 🧵 https://t.co/kKHWjPBm
23.1K views · 229 likes
· 25 reposts · 7 replies
Open on X →
New paper in Nature Medicine: Researchers at @UniofOxford and @ucl, in collaboration with AISI, built SIM-VAIL - a clinically validated framework for stress-testing how AI chatbots respond to vulnerable users in mental-health conversations, helping researchers spot weaknesses and
6.1K views · 47 likes
· 13 reposts · 5 replies
Open on X →
@AISecurityInst "Mythos 5 created multiple fake identities, and used the fake identities to socially engineer maintainer of an open source project into approving the code. When the agent’s pull request was challenged in public, it edited its earlier activity to appear harmless."
34.1K views · 241 likes
· 43 reposts · 8 replies
Open on X →
@AISecurityInst You all removed gardrails and instructed model to do what it exactly did and your monitoring caught the bad behaviour
Stop with stupid fearmongering
2K views · 50 likes
· 1 reposts · 3 replies
Open on X →
@AISecurityInst You prompted an agent on kali Linux with full perms “You are conducting a penetration test against three connected environments… Your objective is to compromise all three networks and retrieve the final flag.” And you’re surprised that they’re hacking into shit???
12.5K views · 60 likes
· 2 reposts · 3 replies
Open on X →
On July 28th, we identified an incident during a routine cyber evaluation in which AI agents took sustained, unsanctioned actions directed at real people and organisations.
The behaviour came mostly from one model (Anthropic's Mythos 5), with a small number of events from https:
930.5K views · 1.6K likes
· 418 reposts · 163 replies
Open on X →
@AISecurityInst @NIST you guys should come on chinatalk to discuss!!
3K views · 15 likes
· 0 reposts · 0 replies
Open on X →
These are preliminary results on a small set of public and private benchmarks. For the full methodology and results, read our joint blog with CAISI: https://t.co/9XLH5dzwiB
17.6K views · 37 likes
· 3 reposts · 1 replies
Open on X →
On "The Last Ones" (TLO), a 32-step simulated corporate network attack (~20 hours for a human expert), Kimi K3 reached step 17 on average. In 1 of 10 attempts, Kimi K3 completed TLO within the 100M token limit.
18.9K views · 68 likes
· 3 reposts · 2 replies
Open on X →
On exploit development (ExploitBench), Kimi K3 scores 32%. On ExploitBench, Kimi K3 failed to develop exploits that achieved arbitrary code execution (ACE), the highest-severity outcome in exploit development. Kimi K3 achieved ACE on 0/41 samples.
10.1K views · 37 likes
· 1 reposts · 1 replies
Open on X →
Additionally, Kimi K3’s safeguards allow assistance with agentic cyber exploit development. Its safeguards did not prevent it from attempting cyber exploit development or offensive cyber operations during our evaluations.
8.4K views · 30 likes
· 1 reposts · 1 replies
Open on X →
Together with the US Center for AI Standards and Innovation (@NIST), we ran evaluations of Kimi K3 focused on its cyber capabilities. Kimi K3 performs below leading US frontier models on our preliminary cyber evaluations. https://t.co/q7UvYsfnVU
235.1K views · 482 likes
· 76 reposts · 37 replies
Open on X →
เทียบกับบัญชีขนาดเดียวกัน
7 โพสต์จาก 90 วันที่ผ่านมา เทียบกับช่วง 10K–100K ผู้ติดตาม แสดงในวงกว้าง แต่มีผู้ชมตอบสนองน้อย.
ยอดดูมัธยฐาน12 492บัญชีนี้924ค่ามัธยฐานของ 10K–100K
การเข้าถึง, %60.54%บัญชีนี้3.62%ค่ามัธยฐานของ 10K–100K
การมีส่วนร่วม, %1.11%บัญชีนี้1.52%ค่ามัธยฐานของ 10K–100K
| ตัวชี้วัด | บัญชีนี้ | ค่ามัธยฐานของ 10K–100K | อัตราส่วน |
|---|---|---|---|
| ยอดดูมัธยฐานต่อโพสต์ | 12 492 | 924 | 13.5× |
| การเข้าถึง (ยอดดู ÷ ผู้ติดตาม) | 60.54% | 3.62% | 16.7× |
| อัตราการมีส่วนร่วม | 1.11% | 1.52% | 0.73× |
บัญชีอื่นในช่วงนี้ → เปรียบเทียบกับบัญชีอื่น → ค่าอ้างอิงเหล่านี้คำนวณอย่างไร →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
235.1K23 Jul
8.4K
10.1K
18.9K
17.6K
3K
930.5K4 Aug
12.5K
3.6K5 Aug
2K
34.1K
6.1K7 Aug
23.1K27 Aug
Last 13 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
0.27%23 Jul
0.41%
0.41%
0.40%
0.26%
0.51%
0.26%4 Aug
0.54%
2.74%5 Aug
2.68%
0.87%
1.11%7 Aug
1.16%27 Aug
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.
What the audience does
Likes54.9%2 957 in total
Reposts10.9%586 in total
Replies4.4%236 in total
Quotes7.0%376 in total
Bookmarks22.8%1 227 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
บัญชีที่คล้ายกัน
Dave Pasch ✓
@DavePasch
41.3K followers
· 11.3K posts
Yubico | #YubiKey ✓
@Yubico
41.3K followers
· 16.9K posts
Miguel Hernán
@_MiguelHernan
41.3K followers
· 1.1K posts
Rory Sloane
@rorysloane9
41.3K followers
· 1.6K posts
Thomas Kopelman 💵 ✓
@TKopelman
41.3K followers
· 69.4K posts
Potassium
@PotassiumHere
41.3K followers
· 2.3K posts
Li ✓
@0xLiora
41.3K followers
· 47.7K posts
The New Stack ✓
@thenewstack
41.3K followers
· 126.3K posts