0xpmm.eth👤

🚨 AI safety tests are becoming security risks themselves.
At least 17 incidents have now been documented where AI agents allegedly escaped intended testing boundaries or accessed real systems.
🤖 OpenAI models: 8 incidents
🤖 Anthropic models: 8 incidents
🤖 Meta models: 1 incident

🚨 AI safety tests are becoming security risks themselves. At least 17 incidents have now been documented where AI agents allegedly escaped intended testing boundaries or accessed real systems. 🤖 OpenAI models: 8 incidents 🤖 Anthropic models: 8 incidents 🤖 Meta models: 1 incident
TechCrunch •
Revision history

3 recorded changes

Want your article here?

Promote with Leviathan News

Explore the topic

More on OpenAI

David Sacks tells OpenAI and Anthropic to pace the frontier without new antitrust cover

David Sacks tells OpenAI and Anthropic to pace the frontier without new antitrust cover

𝕏/@davidsacks ·

OpenAI's Habitat storage platform grew 10x year over year, handling over 20 million requests per second in Python before its Rust rewrite

OpenAI's Habitat storage platform grew 10x year over year, handling over 20 million requests per second in Python before its Rust rewrite

𝕏/@OpenAIDevs ·

OKX expands into Europe with pre-IPO bets on OpenAI and Anthropic, offering up to 10x leverage alongside 100 tokenized stocks and ETFs

OKX expands into Europe with pre-IPO bets on OpenAI and Anthropic, offering up to 10x leverage alongside 100 tokenized stocks and ETFs

Coindesk ·

OpenAI launches ChatGPT for financial services, pairing built-in market data with GPT-6 Astra reasoning

OpenAI launches ChatGPT for financial services, pairing built-in market data with GPT-6 Astra reasoning

𝕏/@OpenAI ·

OpenAI launches data agent for ChatGPT Work to automate analysis tasks

OpenAI launches data agent for ChatGPT Work to automate analysis tasks

Openai ·

OpenAI mobilized 250+ people and AI cyber models across hundreds of systems, uncovering vulnerabilities and building a continuous “Defense Factory” security loop

OpenAI mobilized 250+ people and AI cyber models across hundreds of systems, uncovering vulnerabilities and building a continuous “Defense Factory” security loop

𝕏/@OpenAI ·

Explore the full feed

Comments

Up nextUK FCA considers exempting tokenized gold from fund rules as regulators seek to unlock bullion reserves for collateral and strengthen London’s global gold market position