Writing on AI risk, adversarial testing, and governance from AetherLab. https://aetherlab.co/blog en-US Tue, 13 Oct 2026 16:29:48 GMT PCI SSC's new guidance highlights the growing security challenges of AI in payment environments. For banks, payment processors and financial institutions, the next step is understanding how AI systems behave when instructions are manipulated, sensitive information is exposed or operational boundaries are challenged. https://aetherlab.co/blog/what-pci-s-new-ai-guidance-means-for-payment-institutions https://aetherlab.co/blog/what-pci-s-new-ai-guidance-means-for-payment-institutions Tue, 13 Oct 2026 12:00:00 GMT Alex Georges, PhD Payments Slopsquatting turns invented package names into a possible software supply chain threat. See what the evidence proves and which controls matter. https://aetherlab.co/blog/when-ai-invents-a-software-package https://aetherlab.co/blog/when-ai-invents-a-software-package Wed, 02 Sep 2026 19:50:28 GMT Alex Georges, PhD Deep Tech A concrete 2028 prediction for AI architecture: models will learn topology as a changing latent state alongside geometry, with a clear test. https://aetherlab.co/blog/ai-learned-geometry-next-it-needs-topology https://aetherlab.co/blog/ai-learned-geometry-next-it-needs-topology Wed, 26 Aug 2026 19:51:08 GMT Alex Georges, PhD Future Predictions Fixed tests can overstate language model safety. Release gates should require adaptive evaluation, false positive limits, and renewed monitoring. https://aetherlab.co/blog/when-near-zero-fails-the-real-test https://aetherlab.co/blog/when-near-zero-fails-the-real-test Wed, 05 Aug 2026 14:08:10 GMT Alex Georges, PhD Deep Tech A release manifest, configuration comparison, and observed use review can keep enterprise approval tied to the AI agent that is actually running. https://aetherlab.co/blog/approval-should-expire-when-an-ai-agent-materially-changes https://aetherlab.co/blog/approval-should-expire-when-an-ai-agent-materially-changes Tue, 04 Aug 2026 07:30:38 GMT Alex Georges, PhD Deep Tech Enterprise AI leaders must test systems with combined text and image inputs. Failure to validate multimodal workflows can trigger shutdowns and revenue losses. https://aetherlab.co/blog/tests-that-examine-text-alone-can-miss-multimodal-ai-failures-and-lead-to-costly https://aetherlab.co/blog/tests-that-examine-text-alone-can-miss-multimodal-ai-failures-and-lead-to-costly Fri, 31 Jul 2026 16:01:28 GMT Alex Georges, PhD Deep Tech Stop approving AI suppliers at the company level. Learn why component-based approval is essential to ensure regulatory compliance and avoid blocked AI… https://aetherlab.co/blog/approve-each-ai-component-by-use-and-market-or-local-ai-laws-will-block https://aetherlab.co/blog/approve-each-ai-component-by-use-and-market-or-local-ai-laws-will-block Fri, 17 Jul 2026 15:09:25 GMT Alex Georges, PhD Governance & Policy Mitigate rising governance costs for downloadable AI models by evaluating hardware and hosting contracts alongside software files to ensure smooth deployments. https://aetherlab.co/blog/downloadable-ai-models-will-raise-approval-costs-through-2028-because-a-few https://aetherlab.co/blog/downloadable-ai-models-will-raise-approval-costs-through-2028-because-a-few Thu, 16 Jul 2026 21:37:44 GMT Alex Georges, PhD Future Predictions Illinois SB 315 requires frontier AI developers to provide testable evidence and undergo third-party safety audits starting in 2027, ensuring greater… https://aetherlab.co/blog/illinois-sb-315-makes-frontier-ai-safety-auditable https://aetherlab.co/blog/illinois-sb-315-makes-frontier-ai-safety-auditable Wed, 08 Jul 2026 20:16:06 GMT Alex Georges, PhD Governance & Policy NIST's formal proof shows AI guardrails can't be universally robust, highlighting liability issues and the importance of studying guardrail failure scaling. https://aetherlab.co/blog/read-the-nist-guardrail-proof-as-a-liability-document https://aetherlab.co/blog/read-the-nist-guardrail-proof-as-a-liability-document Wed, 08 Jul 2026 00:39:36 GMT Alex Georges, PhD Deep Tech Consortium-funded AI deployments remove risk controls, leading to public, examiner-documented failures in community banks and health systems by late 2026. https://aetherlab.co/blog/the-capital-brake-just-came-off-mid-market-ai-and-examiners-will-find https://aetherlab.co/blog/the-capital-brake-just-came-off-mid-market-ai-and-examiners-will-find Wed, 08 Jul 2026 00:39:36 GMT Alex Georges, PhD Future Predictions How topological data analysis research from a UCSD PhD became AetherLab's foundation for detecting AI hallucinations through the shape of model activations. https://aetherlab.co/blog/topological-data-analysis-grad-research https://aetherlab.co/blog/topological-data-analysis-grad-research Sat, 22 Nov 2025 12:00:00 GMT Alex Georges, PhD Deep Tech How to detect and prevent LLM hallucinations in production: semantic similarity checks, confidence scoring, adversarial testing, and layered output validation. https://aetherlab.co/blog/preventing-llm-hallucinations-guide https://aetherlab.co/blog/preventing-llm-hallucinations-guide Sun, 08 Jun 2025 12:00:00 GMT Alex Georges, PhD Deep Tech Trust is the currency of AI adoption. A practical framework for enterprise AI trust: quality gates, transparent uncertainty, human oversight, and monitoring. https://aetherlab.co/blog/building-trust-enterprise-ai https://aetherlab.co/blog/building-trust-enterprise-ai Mon, 05 May 2025 12:00:00 GMT Alex Georges, PhD Governance & Policy How to keep multi-agent AI systems reliable: hierarchical validation, inter-agent protocols, conflict resolution, circuit breakers, and real-time metrics. https://aetherlab.co/blog/multi-agent-systems-quality-control https://aetherlab.co/blog/multi-agent-systems-quality-control Fri, 18 Apr 2025 12:00:00 GMT Alex Georges, PhD Deep Tech Slopsquatting turns AI-invented package names into supply-chain attacks. Learn the attack path and use a practical dependency-review checklist to stop it. https://aetherlab.co/blog/slopsquatting-crisis-ai-security https://aetherlab.co/blog/slopsquatting-crisis-ai-security Sat, 15 Mar 2025 12:00:00 GMT Alex Georges, PhD Red-Team Field Notes Lawsuits, regulatory fines, and lost trust: what poor AI quality really costs, and why risk controls and governance must be in place before a crisis hits. https://aetherlab.co/blog/hidden-costs-of-poor-ai-quality https://aetherlab.co/blog/hidden-costs-of-poor-ai-quality Mon, 10 Feb 2025 12:00:00 GMT Alex Georges, PhD Governance & Policy Corporations promise to regulate their own AI. History says they won't. Why voluntary guidelines fail and what meaningful AI governance actually requires. https://aetherlab.co/blog/ai-regulation-self-regulation-illusion https://aetherlab.co/blog/ai-regulation-self-regulation-illusion Mon, 10 Feb 2025 12:00:00 GMT Alex Georges, PhD Governance & Policy Stanford researchers found AI therapy chatbots fail at crisis intervention, validate delusions, and stigmatize mental illness. Why oversight is urgent. https://aetherlab.co/blog/ai-chatbots-mental-health-crisis https://aetherlab.co/blog/ai-chatbots-mental-health-crisis Wed, 22 Jan 2025 12:00:00 GMT Alex Georges, PhD Red-Team Field Notes An 8B model matched a 70B giant at 1% of the cost. Why right-sized, fine-tuned models beat giant ones on cost, latency, reliability, and hallucination rates. https://aetherlab.co/blog/model-minimalism-smaller-sharper-ai https://aetherlab.co/blog/model-minimalism-smaller-sharper-ai Sat, 18 Jan 2025 12:00:00 GMT Alex Georges, PhD Deep Tech 96% of deepfake videos are non-consensual pornography. Inside the crisis of digital nonconsent: the victims, the legal gaps, and how we can fight back. https://aetherlab.co/blog/deepfakes-digital-nonconsent-safeguarding-reality https://aetherlab.co/blog/deepfakes-digital-nonconsent-safeguarding-reality Wed, 18 Dec 2024 12:00:00 GMT Alex Georges, PhD Governance & Policy