This Week in AI & Cybersecurity: Key Global Moves and Breakthroughs From high-stakes talent wars among Big Tech to India's push for indigenous cybersecurity innovation, this week saw significant developments in artificial intelligence and digital defence

From high-stakes talent wars among Big Tech to India's push for indigenous cybersecurity innovation, this week saw significant developments in artificial intelligence and digital defence. Here are five major updates in AI and cybersecurity.

Meta Continues Talent Poaching Spree from Apple to Build AGI

Meta has made two more senior-level hires from Apple's AI team for its newly launched Superintelligence Labs division. Mark Lee has already joined, while Tom Gunter is expected to follow. Both worked under Ruoming Pang, Apple's former head of Foundation Models, who recently accepted a multi-million-dollar compensation package from Meta. This intensifies the rivalry among AI giants, as Meta aggressively builds toward Artificial General Intelligence (AGI), having already brought on board researchers from OpenAI, Anthropic, and Google.

Perplexity AI Hits USD 18 Billion Valuation Amid Funding Surge

Generative AI search startup Perplexity AI has secured an additional USD 100 million in funding, bringing its valuation back to USD 18 billion. This is an extension of an earlier round that valued the company at USD 14 billion. Founded in 2022, Perplexity has experienced rapid valuation jumps, underlining continued investor interest in core internet services reimagined through AI, particularly in competition with traditional search engines like Google.

DeepMind Unveils Gemini Robotics On-Device Model for Real-World AI Tasks

Google DeepMind announced the release of Gemini Robotics On-Device, a foundation model for vision-language-action (VLA) tasks that operates directly on robotic hardware with low latency. Trained using Aloha robots, the model adapts to tasks such as food preparation or playing cards with as few as 50–100 demonstrations. This on-device model achieved over 60 per cent task success rates, marking a notable step toward deploying multimodal AI in real-world physical environments.

IIT Kanpur's C3iHub Launches Cohort VII to Incubate Cybersecurity Startups

The C3iHub at IIT Kanpur has opened applications for its seventh startup incubation cohort, offering funding support of up to INR 30 lakh (approximately USD 35,900) per startup over two years. The initiative targets Indian startups tackling challenges such as mobile forensics, LLM security, and supply chain risks. Following its recent upgrade to a Technology Translational Research Park, C3iHub continues to strengthen India's cybersecurity capabilities through innovation and academic-industry collaboration.

Commenting on this, Prof. Manindra Agrawal, Director, IIT Kanpur said, "At IIT Kanpur, we are committed to advancing India's cybersecurity landscape through cutting-edge research, deep-tech innovation, and impactful entrepreneurship. C3iHub's seventh startup cohort reflects this vision by nurturing ventures working at the forefront of cybersecurity. Recently upgraded as Technology Translational Research Park, C3i hub has been delivering solutions that address emerging national and global challenges, while also contributing to India's self-reliance."

Google's AI Agent 'Big Sleep' Stops Cyber Exploit Before Deployment

Google CEO Sundar Pichai announced that the company's AI agent Big Sleep proactively detected and neutralised a cyber exploit before it could be executed. Though details remain sparse, this marks a shift in cyber defence from reactive to pre-emptive AI action, potentially paving the way for widespread adoption of AI agents in enterprise and national security systems.

"New from our security teams: Our AI agent Big Sleep helped us detect and foil an imminent exploit. We believe this is a first for an AI agent — definitely not the last — giving cybersecurity defenders new tools to stop threats before they're widespread," Pichai posted on X (formerly Twitter).