AI
8 items · 16 articles in this topic
Regulation and product launches both moved. New York's AI safety law took effect while Google, Anthropic and OpenAI each shipped agent or office-style features. VulnCheck's critical view of Anthropic's Glasswing is the day's counterweight to AI-for-security hype.
- 1
New York's AI safety law took effect, requiring major developers such as OpenAI and Anthropic to register, disclose safety practices and report incidents.
- 2
Google made Gemini 4 Carbon generally available after a limited cybersecurity-defender preview. It adds an agent swarm feature.
- 3
VulnCheck reports that Anthropic's Glasswing vulnerability-finding effort overwhelmed maintainers with underwhelming findings after seven months.
- 4
Anthropic added Claude Sheets, Whiteboards and Data artifact types, pitched at non-technical knowledge workers. The Register frames it as a step toward Microsoft Office territory.
- 5
DeepMind released Gemini Robotics 2.5 along with ASIMOV-jailbreak, an open benchmark for testing robot safety against jailbreaks.
- 6
OpenAI reportedly built a kill switch for its research models that can shut down a misbehaving model in about 35 seconds.
- 7
OpenAI added business profiles to ChatGPT search and discovery. The Verge reports it is also beta-launching verified Business Agents for eligible US e-commerce and online service businesses.
- 8
Apple is reportedly preparing a Mac app that makes running local AI models easier.
More in AI · 8 articles›
A developer reports using 48 AI agents on NASA lidar data to identify an unmapped Maya pyramid in 72 hours.
Hacker News (front page)ai-agentsarchaeologylidarFeed excerpt
Article URL: https://pietromiotti.com/latest/maya-site/?maya Comments URL: https://news.ycombinator.com/item?id=50452455 Points: 225 # Comments: 92
Proposes Whiteout, a method to mitigate LLMs memorizing and reproducing personally sensitive information without over-removing knowledge.
arXiv cs.CR (Cryptography & Security)llmprivacyunlearningFeed excerpt
arXiv:2610.02418v2 Announce Type: replace Abstract: Modern large language models (LLMs) are trained on massive, largely unfiltered datasets, including content scraped from nearly every accessible website and user inputs. As a result, LLMs often memorize and reproduce personally sensitive information (PSI) such as birth dates, phone numbers, and home addresses. This leads to significant privacy risks, particularly for high-profile individuals such as executives, politicians, and judges. Existing mitigations largely rely on machine unlearning. However, these methods often remove more information than needed, degrade model utility and safety, and are highly vulnerable to attacks. This paper presents Whiteout, a practical tool that, upon requests by individuals, prevents LLMs from regurgitating their genuine PSIs, by overwriting them using precise and carefully designed obfuscation samples. We evaluate Whiteout on modern LLMs of varying sizes and makers, including a widely-used OpenAI model. Results show that Whiteout effectively prevents disclosure of the targeted PSIs, has negligible impact on model utility and safety, and outperforms existing alternatives. We also test Whiteout agai…
PSI used an unreleased Anthropic model to make progress on a quantum computing question that scientists have pursued for about two decades.
TechCrunchanthropicquantumai-for-scienceFeed excerpt
PSI used an unreleased Anthropic model to answer a fundamental question about quantum computing that scientists have pursued for nearly two decades.
OpenAI added business profiles to ChatGPT, letting businesses be discovered in its consumer search and discovery features.
TechCrunchopenaichatgptllmFeed excerpt
OpenAI is looking to make ChatGPT a more robust consumer discovery engine, and a host of new features seek to empower businesses to help it do that.
Anthropic announced Claude Whiteboards, Claude Sheets, and Claude Data, new features for its Claude assistant.
Hacker News (front page)claudeanthropicllmFeed excerpt
Article URL: https://claude.com/resources/articles/sheets-whiteboards-data Comments URL: https://news.ycombinator.com/item?id=50443716 Points: 96 # Comments: 78
Anthropic unveils an AI-powered office suite of tools aimed at non-technical knowledge workers.
The VergeanthropicllmproductivityFeed excerpt
For years, Anthropic has been known for its focus on enterprise and coding tools that attract high-flying corporate clients, inspiring competing frontier labs - like OpenAI, SpaceXSI, and Meta - to copy its strategies. Earlier this year, Claude Code went through its own particular hype cycle. And now, Anthropic is trying to expand its footprint to other knowledge workers with an AI-powered suite of office tools. At an event in San Francisco on Tuesday, Anthropic unveiled new tools for non-technical workers who spend a lot of time attending meetings, brainstorming new ideas, and analyzing metrics: Claude Meetings, Claude Sheets, and Claude W … Read the full story at The Verge.
Meta's AI assistant Muse is now available on Apple Watch and Wear OS with voice, text, and dictation input.
TechCrunchmetawearablesassistantFeed excerpt
Users can now interact with Meta’s AI assistant through voice calls, text messages, or dictation on their Apple Watch.
Reddit introduces AI-generated product reviews summarizing user opinions on 10,000 products ahead of the holiday shopping season.
TechCrunchredditllmproduct-launchFeed excerpt
Reddit's new AI-powered reviews summarize what users like and dislike about 10,000 products, helping shoppers research purchases ahead of the holidays.