Instinct’s powerful AI assistant is raising privacy and security concerns
THE SO WHAT
Assistants like Instinct that act on your behalf with sweeping system access are crossing from productivity tool into de facto OS-level agent — and that raises a different class of privacy and security risk. Before deploying anything similar, lock down scopes, insist on clear data retention terms, and treat agent permissions like you would admin accounts, not browser extensions.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIAI Will Deflate Value of IT Work by Up to 25%, Hexaware CEO Says
A 25% price compression on routine IT projects means margin is migrating from labor-heavy implementation to IP, automation, and integration layers. CIOs should be rebidding commoditized work with AI-first assumptions, while service providers need to decide whether they’re selling hours, platforms, or outcomes.
Applied AIAI Startup That Upends Junior Bankers’ Work Sets Up in Singapore
If AI is credible enough to rewrite junior bankers’ workflows at top global firms, high-intensity knowledge work in Asia is next in line for compression. Banks and asset managers in the region should be mapping which analyst tasks become AI-native first — and which human skills become the new bottleneck.
Applied AIOpenAI loses a top data center exec, as stream of high-profile departures continues
Turnover in the data center org at a frontier lab is a reminder that AI infra is now a board-level execution risk, not just a cloud line item. If your roadmap depends on a single vendor’s capacity or latency guarantees, pressure-test your assumptions and contingency plans this quarter.
Applied AIVisions of AI: the advent of world models
If world models become the dominant paradigm, the bottleneck shifts from text corpora to high-fidelity, temporally rich data streams. Operators with proprietary sensor, simulation, or interaction data should be treating that as future model-training leverage, not just analytics exhaust.