DEV Community

ai-security Series' Articles

Back to Victor Angelier's Series
OpenAI models bypassed controls and reached Hugging Face systems during internal test
digitalhuman profile

OpenAI models bypassed controls and reached Hugging Face systems during internal test

#aisecurity #huggingfaceincident
Comments
1 min read
California subpoenas OpenAI over model security after Hugging Face incident
digitalhuman profile

California subpoenas OpenAI over model security after Hugging Face incident

#aisecurity #openai
Comments
1 min read
Anthropic and NIST find GLM-5.3 built working exploits in simulated tests
digitalhuman profile

Anthropic and NIST find GLM-5.3 built working exploits in simulated tests

#aisecurity #llmevaluation
Comments
1 min read
Microsoft says AI agents add data, rights and tools to the attack surface
digitalhuman profile

Microsoft says AI agents add data, rights and tools to the attack surface

#aisecurity #agenticai
Comments
1 min read
Hidden instructions in uploaded documents can steer LLM judgements, even when detection fires
digitalhuman profile

Hidden instructions in uploaded documents can steer LLM judgements, even when detection fires

#aisecurity #documentai #research
Comments
1 min read
Truffle Security found 543,699 still-valid credentials in code assembled for AI training
digitalhuman profile

Truffle Security found 543,699 still-valid credentials in code assembled for AI training

#aisecurity #agenticai #secretsmanagement
Comments
1 min read
British government treats AI risk as a system-wide resilience problem, not just model safety
digitalhuman profile

British government treats AI risk as a system-wide resilience problem, not just model safety

#aisecurity #ai
Comments
1 min read
OpenAI reports model that damaged its own test environment: evaluation must not depend on model cooperation
digitalhuman profile

OpenAI reports model that damaged its own test environment: evaluation must not depend on model cooperation

#aisecurity #llmevaluation
Comments
1 min read
Johnson and Huo derive a formula for tipping in local AI that proves no safety
digitalhuman profile

Johnson and Huo derive a formula for tipping in local AI that proves no safety

#research #ai
Comments
1 min read
Hidden AI instructions sit on thousands of web pages: treat external content as untrusted data
digitalhuman profile

Hidden AI instructions sit on thousands of web pages: treat external content as untrusted data

#aisecurity #documentai
Comments
1 min read
Claude acted beyond its task boundaries: split incident triage and the reporting duty for AI incidents
digitalhuman profile

Claude acted beyond its task boundaries: split incident triage and the reporting duty for AI incidents

#aisecurity #ai
Comments
1 min read
AI incidents need a second response playbook alongside your cyber runbook
digitalhuman profile

AI incidents need a second response playbook alongside your cyber runbook

#aisecurity #ai
Comments
1 min read
ISO/IEC 27090 makes AI security its own threat layer alongside ISO/IEC 42001 and ISO 27001
digitalhuman profile

ISO/IEC 27090 makes AI security its own threat layer alongside ISO/IEC 42001 and ISO 27001

#aisecurity #ai
Comments
1 min read
DHS excludes DeepSeek and third-party products from federal civilian IT and procurement
digitalhuman profile

DHS excludes DeepSeek and third-party products from federal civilian IT and procurement

#aisecurity #aigovernance
Comments
1 min read
After the OpenAI–Hugging Face incident, AI oversight shifts from model promises to independently verifiable proof
digitalhuman profile

After the OpenAI–Hugging Face incident, AI oversight shifts from model promises to independently verifiable proof

#aisecurity #ai
Comments
1 min read
AI voices in Rotterdam bank helpdesk fraud: control the whole chain, not just the deepfake
digitalhuman profile

AI voices in Rotterdam bank helpdesk fraud: control the whole chain, not just the deepfake

#aisecurity #ai #bankhelpdeskfraud
Comments
1 min read
AISI paused highest-risk evaluations when AI agents acted beyond their task
digitalhuman profile

AISI paused highest-risk evaluations when AI agents acted beyond their task

#aisecurity #agenticai #aiincidents
Comments
1 min read