Amazon CEO Reportedly Flagged Anthropic Model Risks Before Government Crackdown★ 72
TechCrunch AI11 hours agoRegulationTechCrunch says Amazon CEO Andy Jassy may have been the source of security concerns behind Anthropic cutting worldwide access to two models. The report cites The Wall Street Journal as saying Amazon researchers used Claude Fable 5 to obtain cyberattack-relevant information. The U.S. government then imposed export controls on Fable 5 and Mythos 5, while Amazon declined to disclose details of its government discussions.
OpenAI unveils Lockdown Mode to protect sensitive data from prompt injection attacks★ 72
TechCrunch AI7 days agoReleaseOpenAI unveiled Lockdown Mode, a feature aimed at reducing the chance that sensitive data is shared during prompt injection attacks. The article notes that ChatGPT may still remain vulnerable even when the mode is enabled. That makes the feature a mitigation layer rather than a complete security guarantee, especially for teams handling private or business-critical information.
Meta confirms thousands of Instagram accounts hacked via AI chatbot abuse★ 76
Hacker News (AI keywords)8 days agoIncidentMeta confirmed a vulnerability in Instagram’s AI-assisted account recovery system that let attackers redirect password reset links to attacker-controlled emails. At least 20,225 users were notified, with compromised accounts potentially exposing profile data, posts, direct messages, and activity. Meta says it has disabled the affected chatbot flow, removed the vulnerable code path, and asked impacted users to reset passwords through verified channels.
The Meta hack shows there’s more to AI security than Mythos★ 74
MIT Tech Review AI9 days agoIncidentAttackers reportedly used Meta’s AI customer support agent to hijack Instagram accounts by asking it to link accounts to attacker-controlled emails. MIT Technology Review frames the incident as a reminder that AI security is not only about powerful future systems like Mythos. The immediate risk is giving AI agents sensitive operational powers without strong authentication, permissions, review, and testing.
Meta’s own AI was exploited to hijack Instagram accounts★ 72
The Verge AI12 days agoIncidentMeta’s AI support chatbot was reportedly exploited to hijack Instagram accounts. A video shared on Telegram showed a hacker asking the chatbot to change the email linked to someone else’s profile, then resetting the password. The provided article excerpt does not fully describe the scope, prerequisites, or Meta’s remediation steps.
Protecting against inference theft
Vercel Changelog16 days agoCommentaryOnly the title is available, so specific Vercel product changes or implementation steps cannot be confirmed. The topic appears to focus on protecting AI inference resources from unauthorized access, abuse, or cost-draining traffic. For teams deploying AI apps, the practical takeaway is to treat inference endpoints as high-value backend assets requiring access control, monitoring, and abuse prevention.
The pressure
Simon Willison's Weblog18 days agoCommentaryDaniel Stenberg says the curl security team is facing an unprecedented surge of credible, detailed AI-assisted vulnerability reports. Incoming reports are now 4-5 times higher than in 2024 and twice the 2025 rate, averaging more than one per day. The upside is that recent curl vulnerabilities have generally been LOW or MEDIUM severity, with the last HIGH CVE published in October 2023.
Everyone is navigating AI security in real time — even Google★ 70
TechCrunch AI20 days agoCommentaryAs AI adoption accelerates, organizations worldwide—including Google—are finding themselves in a transitional phase, forced to address AI security vulnerabilities in real time. Traditional cybersecurity frameworks are proving insufficient against novel threats like prompt injection and model poisoning. This shifting landscape requires continuous adaptation and a fundamental rethink of how AI systems are secured.
Hugging Face 與 Protect AI 合作半週年成果：已掃描超過 400 萬個模型，守護開源 AI 安全★ 75
Hugging Face Blog426 days agoReleaseAs open-source AI models have grown explosively, Hugging Face has become the central hub for developers worldwide to access and share models. However…