News, p. 38
News
427 stories · page 38 of 48Vercel ships a production agent that, by default, can barely do anything
The logs read themselves, the root cause surfaces itself, the fix proposes itself - but only with approval. Vercel Agent keeps its own identity, defaults to read-only, and every further action runs through a tightly scoped plan.
Read →The EU will evaluate AI models for cybersecurity before they even reach the market
The European Commission presented its Action Plan on Cybersecurity and Artificial Intelligence. The plan tasks ENISA with drafting a European blueprint for access to advanced AI capabilities in cybersecurity, and builds capacity to evaluate models before they reach the market.
Read →A vulnerability in Adobe ColdFusion scored 10.0 is being actively exploited
CISA added CVE-2026-48282 to the Known Exploited Vulnerabilities catalog over evidence of active exploitation. NVD scores it 10.0 on CVSS - the maximum value. Federal agencies have until 10 July.
Read →Meta's new generator doesn't draw - it works like an agent
Meta introduced Muse Image and Muse Video. Muse Image doesn't translate a prompt straight into a picture - it calls search and coding tools, self-improves its own generations, and gains from more compute spent at request time.
Read →New Horizons woke up after 321 days of sleep, 9.5 billion kilometers out
NASA confirmed that the New Horizons probe has come out of the longest hibernation in its mission and is in good health. The wake-up happened on 23 June; the spacecraft is already starting to transmit science data from the Kuiper Belt.
Read →Chanel turned the Grand Palais into a dark fairy tale for Blazy's second couture collection
Matthieu Blazy showed his second haute couture collection for Chanel at the Grand Palais - a fairy-tale garden with giant beanstalks, drawn from a storybook found in Gabrielle Chanel's apartment.
Read →LeRobot v0.6.0 adds world models - the robot learns by imagining its next move
Hugging Face released LeRobot v0.6.0: policies built on 'world models', new VLA models, reward models to score success, a unified CLI for evaluation, and six simulation benchmarks for robot learning.
Read →FTC: secretly skewing AI answers may be consumer fraud
The US Federal Trade Commission proposed a position: if a company advertises accurate answers but secretly steers the model's output toward another goal, that can be fraud under Section 5. Public comment period through July 31.
Read →The province of Alberta checked 466 million lines of code with Claude in 20 hours
Alberta's Ministry of Technology and Innovation scanned 466 million lines of government code in 20 hours with Claude Code, to find and fix security gaps. Work that classic methods would have ground away at for years.
Read →