|
🍎 Apple bans AI 'nudify' apps
LINK
|
- Apple has updated its App Store rules to ban AI "nudify" apps, now requiring developers to actively stop and remove AI-generated nonconsensual intimate imagery or risk having their apps pulled from the store.
- The revised rule 1.2 names AI content apps and could target tools like xAI's Grok, which Apple threatened to remove earlier this year over nude deepfakes before xAI added safeguards and won approval for an updated version.
- A new guideline 2.5.19 targets a bait-and-switch tactic, warning that developers who hide a nudify feature during review and add it after approval face removal from the App Store and the Apple Developer Program.
|
🔐 AI solves 25-year NSA number mystery
LINK
|
- An Anthropic AI model solved a 25-year-old cryptography puzzle, tracing five mysterious seeds the NSA gave the U.S. government in the late 1990s to an ordinary random-number program fed timestamps from 1997 and 1998.
- An Anthropic employee asked an internal model to investigate without sharing old theories, and in under 30 minutes it matched the seeds to lrand48, an older C library tool, finding P-256's published seed at the 1,842nd candidate.
- The result makes a hidden NSA backdoor in the NIST curves less likely and requires no action from users, though researchers still can't explain why the program skipped some candidates; bitcoin's secp256k1 curve was never involved.
|
⚖️ Anthropic pitches its own AI safety bill
LINK
|
- Anthropic has put forward its own federal AI safety bill, proposing a system of independent evaluators who would check frontier AI models for serious risks and report problems to the Commerce Department.
- Under the plan, evaluators would register with Commerce and could be paid by the labs they review, as long as the money isn't tied to the outcome, and must flag any "imminent enumerated risk" within 72 hours.
- The bill lets Commerce ask a Washington, D.C. federal judge to halt a risky model's release, but bars any other federal agency or officer from suspending a model's development, deployment, or internal use.
|
🪟 Windows 11 is getting a speed boost
LINK
|
- Microsoft says Windows 11 will feel up to 20% faster, with the system reacting more quickly when you click or type across Start, File Explorer, and Edge, plus memory fixes aimed at machines with 8GB of RAM.
- That 20% figure is a best-case number from comparing two builds on how fast Search, Start, and right-click menus respond, so it's not a guarantee across every task you do.
- PCs can now hold up to 2GB more app data in RAM instead of the slower drive, and on 8GB machines Widgets and the news panel stay unloaded until opened, saving over 200MB.
|
📱 Fake WhatsApp app hits Play Store
LINK
|
- A counterfeit version of WhatsApp has slipped onto Google's Play Store and, as of today, still hasn't been pulled from Android's app marketplace despite being flagged by users.
- Published last month under the developer name Agment Assets sallu, the fake listing copies the real app's screenshots and description, and has reached more than 500,000 downloads, often showing up in sponsored search results.
- The imposter appears to exploit Google Play's Early Access program, which turns off public user reviews, making it harder for people to tell the fraudulent listing apart from the genuine WhatsApp.
|
📉 CoreWeave debt fears drag down AI stocks
LINK
|
- A new report lays out how debt problems at CoreWeave, a "neocloud" firm that rents AI computing power, spooked investors and dragged down other AI-related stocks and chipmakers on Thursday.
- CoreWeave has piled up more than $51 billion in debt to build AI infrastructure, and the cost of insuring that debt against default jumped 33% in a week, a level S&P Global calls credit distress.
- CEO Mike Intrator called the market reaction irrational, pointing to a backlog above $100 billion with clients like Microsoft, Meta, and IBM, though analysts warn the firm still needs its promised cash flows to actually arrive.
|
|
