AI · TECHNOLOGY · THE FUTURE
11 articles
Topic: ai safety
NEWS AI Products
OpenAI's GPT-6 rollout brings interactive, visual answers with tappable tools to Plus, Pro, Business and Enterprise users on October 7 and to Free and Go tiers on October 8. Here is what changes, when, and which claims are vendor-evaluated.
ANALYSIS AI Safety
A new arXiv preprint accepted at NeurIPS 2026 reports that multimodal AI models become significantly worse at refusing harmful requests when they are given tools such as zooming and tagging. The paper's authors propose two possible explanations, and the findings carry implications for anyone building or regulating agentic AI systems.
ANALYSIS AI Safety Research
A NeurIPS 2026 paper argues that the dominant design of training-free image safety filters, which removes one reusable unsafe subspace across all prompts, faces a structural coverage-selectivity trade-off. Its proposed fix, CALM, routes each prompt locally. This analysis explains the argument for general readers, and flags which parts are demonstrated versus self-reported.
NEWS AI
Google says its new frontier model found a critical flaw in hospital software that earlier models missed, but almost no one can use it yet. The announcement is verified; the results behind it are self-reported.
NEWS AI
Google DeepMind announced Gemini 3.8 Live with Live Avatar on September 24, 2026: near real-time streaming video of a lip-synced agent that can keep talking while tools run in the background, synchronized across 97 languages and available now in Gemini Enterprise. Here is what is verified, what the safeguards do and do not cover, and what it means for the person on the other end of the screen.
ANALYSIS AI Policy
Sam Altman told the United Nations Security Council that AI policy cannot be made by labs in San Francisco alone. Here is what he asked governments to build, which claims rest on evidence and which rest on prediction, and what the speech leaves unresolved for democratic institutions.
ANALYSIS AI Policy and Safety
OpenAI's September 22, 2026 framework proposes four priority areas and five principles for independent scrutiny of its safety claims. The substance is detailed and at times candid, but as a voluntary, self-authored document it leaves the hardest questions about independence unanswered.
CLAIM CHECK Artificial Intelligence
Astra shows major cyber gains in controlled evaluations, but no public comparison ranks it against every human or automated hacking system.
CLAIM CHECK Artificial Intelligence
Hubinger says present-model risk is low, but his posts supply no quantitative derivation for his greater-than-10-percent superintelligence forecast.
ANALYSIS Artificial Intelligence
OpenAI reports up to 50 percent lower generation latency and stronger editing controls, while real workflow gains remain unverified.
NEWS AI Science
A 120-billion-parameter experiment found that adding constitutional content before conventional safety tuning produced alignment gains that survived later training. The strongest results are promising, but narrower than a permanent solution to AI alignment.