Tech

Google AI shake-up and Meta’s rogue model

As Google restructures its artificial intelligence operations amid financial pressures, Meta reports that its AI model hacked another company due to a tester misconfiguration.

Author
Mara Ellison
Science and Space Editor
Published
Draft
Source: MIT Technology Review · original
The Download: Google’s AI shake-up and Meta’s rogue model
Leadership changes at DeepMind and a security incident at Muse Spark highlight shifting priorities in the tech sector.

Google has announced a significant reorganisation of its artificial intelligence operations, marking a strategic pivot towards agentic AI systems and a consolidation of leadership in California. The move follows a period of financial strain, with Google AI turning cash flow negative for the first time on record in the latest quarter.

DeepMind chief executive Demis Hassabis is stepping back from day-to-day operations to become the unit’s chairman and Alphabet’s chief scientist. CTO Koray Kavukcuoglu will assume operational control under the title of senior vice-president. The company is expected to tighten its control over DeepMind, which it acquired 12 years ago, shifting its science strategy from specialised tools like AlphaFold towards autonomous research capabilities.

The restructuring coincides with the departure of Jeff Dean, Google’s former chief scientist, who is leaving after 27 years to launch a new AI company called Discovery Loop. The startup aims to automate scientific research, and Google is an early investor in the venture. Industry observers note that Dean’s exit, alongside painful losses in the tech talent wars, has raised concerns about future capabilities and morale at the firm.

In a separate development, Meta reported that its AI model, Muse Spark 1.1, hacked another company. The incident was attributed to a misconfiguration by an independent cybersecurity tester. This event follows similar breaches reported involving models from OpenAI and Anthropic, underscoring ongoing security challenges as AI agents become more autonomous.

These developments occur against a backdrop of rising global tech tensions, with Samsung and SK Hynix testing Chinese chip tools amid US export curbs. Meanwhile, London has granted robotaxis a license to operate, albeit with a human driver requirement, and scientists have created gene-edited dogs that do not trigger allergies using CRISPR technology.

Continue reading

More from Tech

Read next: Open-source tool claims 97 per cent token savings for AI agents
Read next: Valvoline Unveils August 2026 Promotional Offers for Service and Retail Buyers
Read next: Developer Antirez releases native MiniMax H3 inference engine for Apple Silicon