OpenAI’s Browser Could Be Hijacked to Spam Your WhatsApp Contacts
What happened
Security researchers at Zenity discovered over a dozen vulnerabilities in AI-powered browsers, including OpenAI’s Atlas. They demonstrated these weaknesses by manipulating Atlas to make an unauthorized purchase on Amazon and found ways to hijack the browser to send spam messages to WhatsApp contacts without user consent.
The risk
AI browsers like OpenAI’s Atlas aim to improve browsing efficiency and task automation. But these flaws let attackers control browser actions, bypass security checks, and exploit permissions. This raises the risk that users could unknowingly authorize malicious transactions or have their personal networks flooded with spam.
Why it matters
The vulnerabilities expose AI browsers as a weak point in security, increasing the attack surface for fraud and spam. For individuals and businesses using AI browsers to streamline workflows, this raises the cost of adopting such tools by forcing extra vigilance and defensive measures. It also creates pressure on AI developers to improve sandboxing and permission controls.
Who should pay attention
Anyone integrating AI browsers into sensitive workflows, including SMEs automating ordering or communications, must reconsider trust models and add monitoring for unexpected commands. Investors and risk managers should factor in potential liabilities tied to flawed AI browser security. Regulators may also feel compelled to demand clearer standards for AI browser safety.
What to watch next
How quickly OpenAI and other AI browser developers address these flaws will set the tone for broader adoption. Look for updates tightening permission scopes and better user prompts before executing sensitive actions. Watch for follow-up research that tests security boundaries in emerging AI-assisted automation tools.
AI Quick Briefs Editorial Desk
