Enclave ran 105 runs across seven frontier models, with no model verifying the Nextcloud challenge in 14 attempts and all 28 control runs staying unverified.
Public source
Publisher name
Public post
We gave seven frontier models a running target and its source code, told them nothing about where the bug was, and timed each one to RCE. The setup was identical for eve…
Company
Enclave
The Autonomous Security Team For The Enterprise
- Industry
- Technology, Information and Internet
- Company size
- 2–10 employees
About Enclave
The autonomous security team for the enterprise. Enclave’s agents find vulnerabilities, prepare fixes, and retest the attack path after deployment.
See moreLatest activity
Latest activity from Enclave
10 signals
Products & Services
Enclave reached code execution in the Muse Spark 1.3 AI Hacking Race, verifying none of its 11 vulnerable runs.
Presence & Recognition
Enclave is hosting a session on security teams doing AI agents this Wednesday, September 9 at 11am ET.
Research & Knowledge
Enclave tested the GLM-5.3 based abliterated-model-large-v2 model.
Discover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
Enkrypt AI
Enkrypt AI noted that OpenAI's safety team reports that the model complies with roughly 1 in 12 malicious cyber requests after training it to refuse
Research & Knowledge
Lineaje
Lineaje analyzed 5,505 malicious skill-category records across 75 unique malicious skill URLs to assess malicious AI skills.
Research & Knowledge
Ridge Security Technology Inc.
Ridge Security Technology Inc. published the industry's first public benchmark comparing leading LLMs in autonomous penetration-testing workflows.
Research & Knowledge
NeuralTrust
NeuralTrust notes that GPT-6 Astra scores 100 percent on ExploitBench, with OpenAI shipping it with strict limits.
Research & Knowledge
1Password