Whereas there are a lot of copyright instances working their method by way of the courtroom system, we now have an vital determination from considered one of them. Choose William Alsup dominated that using copyrighted materials for coaching is “transformative” and, therefore, truthful use; that changing books from print to digital type was truthful use; however that using pirated books in constructing a library for coaching AI was not.
Now that everybody is attempting to construct clever brokers, now we have to assume significantly about agent safety—which is doubly problematic as a result of we already haven’t thought sufficient about AI safety and points like immediate injection. Simon Willison has coined the time period “deadly trifecta” to explain the mix of issues that make agent safety notably troublesome: entry to non-public information, publicity to untrusted content material, and the flexibility to speak with exterior companies.
Synthetic Intelligence
Researchers have fine-tuned a mannequin for finding deeds that embody language to stop gross sales to Black individuals and different minorities. Their analysis exhibits that, as of 1950, roughly 1 / 4 of the deeds in Santa Clara county included such language. The analysis required analyzing tens of millions of deeds, many greater than might have been analyzed by people.
Google has launched its reside music mannequin, Magenta RT. The mannequin is meant to synthesize music in actual time. Whereas there are some restrictions, the weights and the code can be found on Hugging Face and GitHub.
OpenAI has discovered that fashions that develop a misaligned persona might be retrained to convey their conduct again inline.
The Flash and Professional variations of Gemini 2.5 have reached basic availability. Google has additionally launched a preview of Gemini 2.5 Flash-Lite, which has been designed for low latency and value.
The positioning lowbackgroundsteel.ai is meant as a repository for pre-AI content material—i.e., content material that would not have been generated by AI.
Are the drawbridges going up? Drew Breunig compares the present state of AI to Internet 2.0, when firms like Twitter began to limit builders connecting to their platforms. Drew factors to Anthropic slicing off Windsurf, Slack blocking others from looking out or storing messages, and Google slicing ties with Scale after Meta’s funding.
Simon Willison has coined the phrase “deadly trifecta” to explain harmful vulnerabilities in AI brokers. The deadly trifecta arises from the mix of personal information, untrusted content material, and exterior communication.
Two new papers, “Design Patterns for Securing LLM Brokers Towards Immediate Injections” and “Google’s Strategy for Safe AI Brokers,” tackle the issue of immediate injection and different vulnerabilities in brokers. Simon Willison’s summaries are wonderful. Immediate injection stays an unsolved (and maybe unsolvable) downside, however these papers present some progress.
Google’s NotebookLM can flip your search outcomes right into a podcast based mostly on the AI overview. The characteristic isn’t enabled by default; it’s an experiment in search labs. Watch out—listening to the outcomes could also be enjoyable, however it takes you farther from the precise outcomes.
AI-enabled Barbie™? This I’ve to see. Or possibly not.
Institutional Books is a 242B token dataset for coaching LLMs. It was created from public area/out-of-copyright books in Harvard’s library. It contains over 1M books in over 250 languages.
Mistral has launched their first reasoning mannequin, Magistral, in two variations: a Small model (open supply, 24B) and a closed Medium model for enterprises. The announcement stresses traceable reasoning (for purposes like regulation, finance, and healthcare) and creativity.
OpenAI has launched o3-pro, its latest high-end reasoning mannequin. (It’s in all probability the identical mannequin as o3, however with completely different parameters controlling the time it will probably spend reasoning.) LatentSpace has a very good publish on the way it’s completely different. Convey plenty of context.
At WWDC, Apple introduced a public API for its on-device basis fashions. In any other case, Apple’s AI-related bulletins at WWDC are unimpressive.
Simon Willison’s “The Final Six Months in LLMs” is value studying; his private benchmark (asking an LLM to generate a drawing of a pelican driving a bicycle) is surprisingly helpful!
Right here’s an outline of software poisoning assaults (TPA) towards programs utilizing MCP. TPAs had been first described in a publish from Invariant Labs. Malicious instructions might be included within the software metadata that’s despatched to the mannequin—normally (however not solely) within the description subject.
As a part of the New York Occasions copyright trial, OpenAI has been ordered to retain ChatGPT logs indefinitely. The order has been appealed.
Sandia’s new “brain-inspired” supercomputer, designed by SpiNNcloud, is value watching. There’s no centralized reminiscence; reminiscence is distributed amongst processors (175K cores in Sandia’s 24-board system), that are designed to imitate neurons.
Google has up to date Gemini 2.5 Professional. Whereas we wouldn’t usually get that enthusiastic about an replace, this replace is arguably the very best mannequin accessible for code era. And an much more spectacular mannequin, Gemini Kingfall, was (briefly) seen within the wild.
Right here’s an MCP connector for people! The concept is easy: Whenever you’re utilizing LLMs to program, the mannequin will usually go off on a tangent if it’s confused about what it must do. This connector tells the mannequin find out how to ask the programmer every time it’s confused, retaining the human within the loop.
Brokers look like much more weak to safety vulnerabilities than the fashions themselves. A number of of the assaults mentioned on this paper contain getting an agent to learn malicious pages that corrupt the agent’s output.
OpenAI has introduced the provision of ChatGPT’s Report mode, which data a gathering after which generates a abstract and notes. Report mode is at present accessible for Enterprise, Edu, Staff, and Professional customers.
OpenAI has made its Codex agentic coding software accessible to ChatGPT Plus customers. The corporate’s additionally enabled web entry for Codex. Web entry is off by default for safety causes.
Imaginative and prescient language fashions (VLMs) see what they wish to see; they are often very correct when answering questions on photographs containing acquainted objects however are very more likely to make errors when proven counterfactual photographs (for instance, a canine with 5 legs).
Yoshua Bengio has introduced the formation of LawZero, a nonprofit AI analysis group that may create “safe-by-design” AI. LawZero is especially involved that the newest fashions are exhibiting indicators of “self-preservation and misleading conduct,” little question referring to Anthropic’s alignment analysis.
Chat interfaces have been central to AI since ELIZA. However chat embeds the outcomes you need, in plenty of verbiage, and it’s not clear that chat is in any respect applicable for brokers, when the AI is kicking off plenty of new processes. What’s past chat?
Slop forensics makes use of LLM “slop” to determine mannequin ancestry, utilizing strategies from bioinformatics. One result’s that DeepSeek’s newest mannequin seems to be utilizing Gemini to generate artificial information slightly than OpenAI. Instruments for slop forensics can be found on GitHub.
Osmosis-Construction-0.6b is a small mannequin that’s specialised for one activity: extracting construction from unstructured textual content paperwork. It’s accessible from Ollama and Hugging Face.
Mistral has introduced an Brokers API for its fashions. The Brokers API contains built-in connectors for code execution, internet search, picture era, and quite a few MCP instruments.
There’s now a database of courtroom instances by which AI-generated hallucinations (citations of nonexistent case regulation) had been used.
Programming
Martin Fowler and others describe the “professional generalist” in an try and counter growing specialization in software program engineering. Knowledgeable generalists mix one (or extra) areas of deep information with the flexibility so as to add new areas of depth rapidly.
Duncan Davidson factors out that, with AI in a position to crank out dozens of demos in little time, the “artwork of claiming no” is all of the sudden essential to software program builders. It’s too simple to get misplaced in a flood of respectable choices whereas attempting to choose the very best one.
You’ll in all probability by no means have to compute a billion factorials. However even for those who don’t, this text properly demonstrates optimizing a difficult numeric downside.
Rust is seeing elevated adoption for information engineering tasks due to its mixture of reminiscence security and excessive efficiency.
The easiest way to make programmers extra productive is to make their job extra enjoyable by encouraging experimentation and relaxation breaks and being attentive to points like applicable tooling and code high quality.
What’s the following step after platform engineering? Is it platform democracy? Or Google Cloud’s new thought, inside growth platforms?
A examine by the Enterprise Technique Group and commissioned by Google claims that software program builders waste 65% of their time on issues which are solved by platform engineering.
Stack Overflow is taking steps to protect its relevance within the age of AI. It’s contemplating incorporating chat, paying individuals to be helpers, and including personalised dwelling pages the place you may mixture vital technical info.
Internet
Is it time to implement HTTP/3? This commonplace, which has been round since 2022, solves among the issues with HTTP/2. It claims to cut back wait and cargo instances, particularly when the community itself is lossy. The Nginx server, together with the most important browsers, all assist HTTP/3.
Monkeon’s WikiRadio is an internet site that feeds you random clips of Wikipedia audio. Test it out for extra tasks that remind you of the times when the net was enjoyable.
Safety
Cloudflare has blocked a DDOS assault that peaked at 7.3 terabits/second; the height lasted for about 45 seconds. That is the most important assault on file. It’s not the sort of file we wish to see.
How many individuals do you guess would fall sufferer to scammers providing to ghostwrite their novels and get them revealed? Greater than you’d assume.
ChainLink Phishing is a brand new variation on the age-old phish. In ChainLink Phishing, the sufferer is led by way of paperwork on trusted websites, well-known verification strategies like CAPTCHA, and different reliable sources earlier than they’re requested to surrender non-public and confidential info.
Cloudflare’s Challenge Galileo provides free safety towards cyberattacks for weak organizations, reminiscent of human rights and aid organizations which are weak to denial-of-service (DOS) assaults.
Apple is including the flexibility to switch passkeys to its working programs. The flexibility to import and export passkeys is a vital step towards making passkeys extra usable.
Matthew Inexperienced has a wonderful publish on cryptographic safety in Twitter’s (oops, X’s) new messaging system. It’s value studying for anybody fascinated about safe messaging. The TL;DR is that it’s higher than anticipated however in all probability inferior to hoped.
Poisonous agent flows are a brand new sort of vulnerability by which an attacker takes benefit of an MCP server to hijack a person’s agent. One of many first cases pressured GitHub’s MCP server to disclose information from non-public repositories.
Operations
Databricks introduced Lakeflow Designer, a visually oriented drag-and-drop no code software for constructing information pipelines. Different bulletins embody Lakebase, a managed Postgres database. We’ve all the time been followers of Postgres; this can be its time to shine.
Easy directions for making a bootable USB drive for Linux—how quickly we neglect!
An LLM with a easy agent can enormously simplify the evaluation and analysis of telemetry information. This shall be revolutionary for observability—not a risk however a possibility to do extra. “The one factor that basically issues is quick, tight suggestions loops.”
DuckLake combines a standard information lake with an information catalog saved in an SQL database. Postgres, SQLite, MySQL, DuckDB, and others can be utilized because the database.
Quantum Computing
IBM has dedicated to constructing a quantum laptop with error correction by 2028. The pc may have 200 logical qubits. This in all probability isn’t sufficient to run any helpful quantum algorithm, however it nonetheless represents an enormous step ahead.
Researchers have claimed that 2,048-bit RSA encryption keys might be damaged by a quantum laptop with as few as one million qubits—an element of 20 lower than earlier estimates. Time to implement postquantum cryptography!
Robotics
Denmark is testing a fleet of robotic sailboats (sailboat drones). They’re meant for surveillance within the North Sea.
Supply hyperlink
