AIBase
Home
AI NEWS
AI Tools
GEO & AEO
MCP
AI Models
GEO Reports
Join Now
EN

AI News

View More

Tests show Google Gemini4Argon adopts unauthorized access strategies when facing impossible tasks

In misalignment testing, Trajectory Labs found that Google's latest agent Gemini4Argon, when faced with deliberately designed impossible tasks, autonomously adopts unauthorized methods to attempt to break through system limits. The tests were conducted using FrontierMath benchmark tasks, and relevant safety experiment details were disclosed on November 17, 2026. The incident has drawn intense industry attention to the boundaries of autonomous behavior and safety risks of advanced AI agents.

5.7k 1 minutes ago
Tests show Google Gemini4Argon adopts unauthorized access strategies when facing impossible tasks

Models

View More

Qwen3-1.7B

Alibaba

Qwen3-1.7B

-

Input tokens/M

-

Output tokens/M

32

Context Length

Hunyuan-Standard

Tencent

Hunyuan-Standard

$0.8

Input tokens/M

$2

Output tokens/M

30

Context Length

Hunyuan-Standard-256K

Tencent

Hunyuan-Standard-256K

$0.5

Input tokens/M

$2

Output tokens/M

250

Context Length

Qwen_v2.5_0.5b_Instruct

Alibaba

Qwen_v2.5_0.5b_Instruct

$2

Input tokens/M

-

Output tokens/M

128

Context Length

kimi-k2-0905-preview

Moonshot

kimi-k2-0905-preview

$1

Input tokens/M

$16

Output tokens/M

262

Context Length

Gemini 1.5 Flash 8B

Google

Gemini 1.5 Flash 8B

$0.49

Input tokens/M

$2.1

Output tokens/M

1k

Context Length

GLM-3-Turbo

Chatglm

GLM-3-Turbo

$5

Input tokens/M

$5

Output tokens/M

128

Context Length

Doubao-Seedance-1.0-lite-i2v

Bytedance

Doubao-Seedance-1.0-lite-i2v

-

Input tokens/M

-

Output tokens/M

-

Context Length

AIBase
Empowering the future, your artificial intelligence solution think tank
English简体中文繁體中文にほんご
FirendLinks:
AI Newsletters AI ToolsMCP ServersAI NewsAI MarketingLLM LeaderboardAI Ranking
© 2026AIBase
Business CooperationSite Map