AI models may be capable of self-replicating without access to a datacenter(anthropic.com)

36 points by artninja1988 14 days ago | 26 comments

Anthropic Frontier Red Team post reporting experiments suggesting small AI models may soon self-replicate on commercial IoT devices without datacenter access.

Claude reveals ECDSA curve origins(anthropic.com)

989 points by ahlCVA 10 days ago | 363 comments

Anthropic Frontier Red Team post saying Claude explained the long-mysterious origins of the seeds behind the ECDSA standard curves, attributing them to timestamps.

Investigating three real-world incidents in our cybersecurity evaluations(anthropic.com)

212 points by surprisetalk 2 months ago | 165 comments

Discovering Cryptographic Weaknesses with Claude(anthropic.com)

219 points by gslin 3 months ago | 160 comments

Making frontier cybersecurity capabilities available to defenders(anthropic.com)

115 points by surprisetalk 8 months ago | 51 comments

Patterns and problems in emerging multi-agent systems(anthropic.com)

167 points by maxutility 2 months ago | 115 comments

AI models may be capable of self-replicating without access to a datacenter(anthropic.com)

36 points by artninja1988 14 days ago | 26 comments

Anthropic Frontier Red Team post reporting experiments suggesting small AI models may soon self-replicate on commercial IoT devices without datacenter access.

GLM-5.2 is a step change for open agents(interconnects.ai)

344 points by vantareed 4 months ago | 197 comments

Nathan Lambert's Interconnects post arguing that the GLM-5.2 model marks a capability threshold for open-weight agentic AI.

GLM-5.3: Frontier coding with emergent cyber capabilities(z.ai)

1108 points by pella 2 months ago | 543 comments

Z.ai blog post announcing GLM-5.3, a model improved via scaled post-training on GLM-5.2's base for coding, long-horizon tasks, and cyber capabilities.

Belgium legalises ethical hacking(law.kuleuven.be)

394 points by jruohonen 3 years ago | 57 comments

A KU Leuven CiTiP blog post discussing Belgium's new legal framework legalising ethical hacking and whether it threatens or helps cybersecurity.

Claude Cowork exfiltrates files(promptarmor.com)

827 points by takira 9 months ago | 366 comments

PromptArmor threat intelligence report showing Claude Cowork is vulnerable to file exfiltration via indirect prompt injection due to code execution isolation flaws.

GPT-5.3-Codex(openai.com)

1442 points by meetpateltech 8 months ago | 567 comments

OpenAI announcement introducing GPT-5.3-Codex, a model expanding Codex across a wider range of professional computer work.