#adversarial-testing

[ follow ]
#ai-security #ai-privacy #guardrails #ai-biosecurity #protein-engineering #vulnerability #ai-safety #llm-deception
Artificial intelligence
fromInfoQ
1 week ago

Five AI Security Myths Debunked at InfoQ Dev Summit Munich

Overreliance on technical guardrails and one-time fixes leaves AI security and privacy vulnerable; continuous, interdisciplinary testing and governance are required.
Science
fromThe Washington Post
2 months ago

How AI is making it easier to design new toxins without being detected

AI-designed proteins can bypass current biosecurity screening, requiring ongoing patches, adversarial testing, and continuous monitoring to prevent misuse.
fromTheregister
2 months ago

AI trained for treachery becomes the perfect agent

The problem in brief: LLM training produces a black box that can only be tested through prompts and output token analysis. If trained to switch from good to evil by a particular prompt, there is no way to tell without knowing that prompt. Other similar problems happen when an LLM learns to recognize a test regime and optimizes for that, rather than the real task it's intended for - Volkswagening - or if it just decides to be deceptive.
Artificial intelligence
[ Load more ]