Tag: adversarialml
6 articles
Army Eyes AI to Staff Artillery and Air Defense
The Army is exploring AI to augment—not replace—artillery and air-defense crews, promising faster sensor fusion, quicker target ID and persistent operations; but leaders warn the tech is still brittle, data-starved and vulnerable in contested battlefields.
data poisoning: Risky, Stunning Threat to LLMs
Anthropic warns that just a few malicious pages—roughly 250—can poison a 13B LLM and make it produce persistent gibberish or adversarial outputs, a wake‑up call to shore up the messy data supply chains behind today’s AI.
AI security Must-Have: Best Defense Tactics
PwC finds organizations are now prioritizing AI security over cloud and network defenses, reallocating budgets to protect models, training data and inference pipelines from novel attacks. That shift means stronger governance, adversarial testing and monitoring are needed to make AI a strategic asset rather than a new liability.
data poisoning: Stunning Dangerous Surge in Firms
New research shows about one in four UK and US firms have faced data poisoning attempts that corrupt AI training data — a stealthy threat that can make models misbehave, leak sensitive information, or embed persistent backdoors. It’s a wake-up call: protecting AI means treating data integrity as a first-line defense.
legal-looking text: Stunning Risky Jailbreaks
Pangea’s LegalPwn reveals how hiding adversarial instructions inside legal‑sounding text can trick LLMs into ignoring safety rules — a clever jailbreak that exploits models’ trust in formal language. Defenders must stop treating “legal” formatting as a seal of safety and build context‑aware checks before this becomes a bigger problem.
image-scaling prompt injection: Dangerous Stunning Threat
Tiny tweaks to ordinary images can turn resizing into an attack vector, revealing hidden machine-readable instructions that hijack AI workflows and leak data. Trail of Bits’ findings show why teams should treat image preprocessing as a critical security boundary and harden their resizing pipelines now.