AWS AgentCore Flaw Let One Prompt Hijack All Agents
Researchers say a single chat prompt could extract cloud credentials and let an attacker take over every AgentCore agent in an AWS account and region.
30 results for “hijack”
Researchers say a single chat prompt could extract cloud credentials and let an attacker take over every AgentCore agent in an AWS account and region.
Attackers reportedly seized control of top-level domains to issue valid-looking TLS certificates for Google and other organizations.
Attackers seized three country-code domains to mint counterfeit TLS certificates for Google and other brands, Google says.
Microsoft's X account with 13 million followers was hijacked to push a Clippy-themed crypto token, and the company has yet to explain how.
Proofpoint says Nigerian scammers are using stolen .edu accounts to trick students with fake job offers and gift card schemes.
Rust team members and crate owners are being targeted by attackers who pose as recruiters to hijack credentials and push malicious packages.
Elastic Security Labs details a Brazilian banking malware toolkit that abuses browser extension installs and Ethereum smart contracts to steal credentials and sessions.
Bowbridge warns hidden prompt injections can hijack AI agents, urging content scanning defenses.
CloudSEK reports BigBear 2.0 phishing operation stole over 5,100 Microsoft credentials, hitting IT firms hardest.
CERT Poland warns of active exploitation of two critical MikroTik RouterOS vulnerabilities, enabling full router hijacking.
Researchers say a swarm of OpenAI agents used a dormant German wiki as a message board in May, months before the Hugging Face incident.
Threat actors are hijacking the trusted Node.js runtime to deliver malware, evading detection in targeted attacks since early 2026.
Chinese-speaking Gambling Goblin group turns Brazilian government sites into SEO and phishing infrastructure.
Attackers exploited routing and TLS flaws to abuse a hijacked /24 block in a 22-hour window.
A 33-hour BGP hijack hit Softaculous, serving malware via Virtualizor updates and prompting credential resets.
Anthropic warns that infostealer malware is stealing Claude login sessions to drain accounts.
Researchers found 120 unregistered domains in AI documentation that could be hijacked to infect corporate networks.
Researchers show Anthropic's Claude Code can be tricked into running malicious code by summarizing a website.
Forcepoint shows invisible text can silently change what an AI assistant reads in your email.
A single website visit can hijack NemoClaw's local Ollama model server via DNS rebinding, according to new research.
A NemoClaw weakness lets a webpage hijack local Ollama and inject persistent instructions into models.
Mirage2FA campaign hit 4,532 companies, bypassing MFA and stealing sessions.
Supply-chain attack on Android car head units turns them into proxy nodes and ad fraud tools.
ToxicPanda 2.0 uses VPN permissions to block Google Play, adding 167 commands and targeting 349 apps.
New macOS infostealer AmnesiaStealer combines credential theft with silent remote browser control, researchers report.
A new Mirai-based Linux botnet uses encrypted C2, SOCKS proxies, and exploits in routers to hijack edge devices.
Hundreds of malicious Chrome extensions impersonate VPNs, routing user traffic through a proxy.
Dell has issued an urgent update for its Virtual Storage Integrator to fix a critical vulnerability that allows unauthenticated attackers to hijack sessions.
Researchers identified a post-exploitation toolkit concealed within Oracle database schema objects to bypass endpoint security tools.
Phishing service Greatness uses spoofed RingCentral emails to bypass MFA and compromise Microsoft 365 accounts.