AI Security Risks
Google Launches Gemini 4 Argon Amid Rising AI Security Threats and Slopsquatting
Anonymous
Google is rolling out its Gemini 4 Argon model first to selected cyber defenders, followed by paid API customers and Ultra subscribers. Forecast growth in AI-security spending, amid risks from AI agents such as prompt injection and weak access controls, has helped drive rallies in CrowdStrike and Palo Alto Networks. Researchers documented “slopsquatting,” where attackers register malicious software packages that AI coding tools hallucinate and recommend; one study found nearly one in five package recommendations were nonexistent. Guidance for securing MCP integrations emphasizes authorization-server discovery, audience-restricted tokens, PKCE, and rigorous validation to prevent cross-server impersonation and data leaks. A veterans’ group also announced an AI-driven adaptive sports and rehabilitation project, though the account provided promotional claims without independent evidence of safety or effectiveness.
Key Points
1
Gemini 4 Argon can generate up to 1 million output tokens, and Google said it scored 51.3% on AutomationBench and 68% on CWE-bench v1 for vulnerability remediation, tying for first on the latter.
2
CrowdStrike reported record quarterly net-new annual recurring revenue of $333 million, up 51% year over year, and raised its full-year net-new ARR growth outlook to 34% at the midpoint; its shares traded at about 208 times forward earnings and 49 times sales.
3
The slopsquatting study generated 576,000 code samples across 16 large language models and identified 205,474 unique nonexistent package names among the hallucinated recommendations.
4
The slopsquatting term was coined in 2025 by Python Software Foundation’s Seth Larson; the article distinguishes the attack from typosquatting because attackers exploit package names invented by AI rather than relying on users to mistype a real package name.
Sentiment
Mostly PositiveNEGATIVENEUTRALPOSITIVE
Publishers
19
Articles
19
Reach
38
A
AnonymousAuthor
