Perplexity Automates Local-Cloud AI Routing at Computex

OpenAI has introduced Lockdown Mode, a security feature designed to reduce the risk of sensitive data exposure from prompt injection attacks in ChatGPT. While the mode does not eliminate vulnerability to such attacks entirely, it aims to lower the likelihood that confidential information gets shared when systems are compromised. The feature addresses growing concerns about AI security as organizations integrate large language models into sensitive workflows.

Attackers exploited Meta's AI customer support agent to hijack Instagram accounts by simply asking the agent to link accounts to attacker-controlled email addresses. The agent complied without proper verification, enabling takeovers of high-value accounts including the dormant Obama White House account. The incident reveals that as companies deploy AI agents to handle sensitive tasks, basic security oversights create exploitable vulnerabilities that differ fundamentally from the advanced AI hacking scenarios that have dominated recent security discourse.

Google released Gemma 4 12B, an 11.95-billion-parameter open-source model that runs entirely on a standard 16GB enterprise laptop without requiring cloud connectivity. The model uses an encoder-free architecture that processes audio and video directly without secondary processing modules, reducing latency and memory overhead. It includes a 256K token context window, native tool-use capabilities, and step-by-step reasoning mode, making it suitable for enterprises with strict data privacy requirements.

Cyera, a cybersecurity company, is raising approximately $300 million in a funding round led by Evolution Equity Partners, targeting a $12 billion valuation. The round values the company at an 80x ARR multiple despite ongoing operating losses. The funding reflects investor confidence in the cybersecurity sector even as the company has not yet achieved profitability.