#backdoor-attacks

[ follow ]
#data-poisoning #model-robustness #antigravity #security-vulnerability #ai-agents #ai-security
Software development
fromInfoWorld
2 weeks ago

Security researchers caution app developers about risks in using Google Antigravity

Google's Antigravity tool contains vulnerabilities that can allow attackers to backdoor compromised workspaces and execute arbitrary code; Google is investigating and will post known issues.
#data-poisoning
fromFuturism
1 month ago
Artificial intelligence

Researchers Find It's Shockingly Easy to Cause AI to Lose Its Mind by Posting Poisoned Documents Online

Posting as few as 250 poisoned documents online can backdoor AI models, enabling trigger-phrase manipulation and creating serious security risks.
fromTechzine Global
2 months ago
Artificial intelligence

Small amount of poisoned data can influence AI models

Approximately 250 poisoned documents can create effective backdoors in LLMs regardless of model size or total training data volume.
fromFuturism
1 month ago
Artificial intelligence

Researchers Find It's Shockingly Easy to Cause AI to Lose Its Mind by Posting Poisoned Documents Online

more#data-poisoning
[ Load more ]