← The Polylog AI Intelligence Brief
Morning Edition · Friday, June 19, 2026
'FloatDoor' Shows How a Backdoor Can Hide in the Deployment Platform, Not the Weights
New security research describes platform-triggered backdoors that change a large language model's outputs based on where it runs, a risk for code-generating models.
A paper titled FloatDoor: Platform-Triggered Backdoors in LLMs examines a threat that is harder to detect than the classic data-poisoning backdoor. The abstract notes that an identical large language model (LLM) can produce measurably diffe…
Continue the AI Intelligence Brief
Track frontier labs, chips, export controls, model releases, regulation, and AI infrastructure.
- 5 AI intelligence signals a day
- Frontier labs, compute, and chips
- Model releases and AI infrastructure
- Source-grounded analysis with confidence labels
The Global Intelligence Brief stays free.
More from this edition
- OpenAI Pushes GPT-5.5 Instant Into Clinical Territory, Claiming Reasoning-Model Health Scores
- DeepSeek Reportedly Raises Over $7.4 Billion in an Unusual First Outside Round
- US Export Directive Suspends All Access to Anthropic's Fable 5 and Mythos 5
- DeepMind Publishes an 'AI Control' Roadmap for Supervising Advanced Agents
- A New Experimental Survey Weighs Diffusion Language Models Against Autoregression
- Researchers Propose Deontic Policies to Govern Agentic AI at Runtime
- Estonia Plans Digital IDs for Autonomous AI Agents
- FERC Action on Large-Load Interconnection Targets the Grid Bottleneck Behind AI Data Centers
- OpenAI Adds Spend Controls and Usage Analytics for Enterprise Customers
- Study Separates Language Relatedness From Task Alignment in Cross-Lingual Transfer
- AI-Generated Content Is Outpacing Human Review, The Economist Reports