Discover the Best AI Tools

We test the top AI tools for writing, video, images & music — so you don't have to.

Explore AI Tools

Anthropic's AI Found 129,000 Software Bugs — Now It's Arming More Security Teams

-

Anthropic just made a big bet on AI-powered defense. It's expanding access to its most powerful AI models — with fewer safeguards — to more vetted cybersecurity teams. The reason: its AI already found over 129,000 real software vulnerabilities this year.


What Project Glasswing uncovered


Partners under Anthropic's Glasswing initiative found at least 129,000 verified software vulnerabilities between April and July 2026, more than 33,000 of them rated critical or high severity. Anthropic's own open-source scanning added 5,500 more between April and October. The company says the true number is likely at least five times higher, since the data covers only a limited group of partners.


The revamped Cyber Verification Program


Anthropic announced Tuesday a revamped Cyber Verification Program (CVP) combining two programs it ran for six months. Glasswing gave critical-software organizations access to Claude Mythos, its most cyber-capable model family, while the original CVP gave vetted teams reduced safeguards on Claude Opus and Sonnet models. The new program has three tiers.


The Defense tier covers incident response and malware analysis for security teams, critical-infrastructure operators, open-source maintainers, and researchers with a track record of reported vulnerabilities. The Red Team tier adds authorized penetration testing and is open to organizations only. The Specialized tier, with the fewest restrictions, is reserved for a small group authorized to test safety-critical systems like power grids, flight systems, and interbank transfer infrastructure. All tiers include access to Claude Opus 5.5, Sonnet 5.5, Mythos 5.1, and future models.


Why it matters


When Anthropic unveiled Claude Mythos Preview in April, fears surfaced that AI could hack software before it could be secured. Anthropic is betting that giving the best tools to defenders first wins that race — and the trend is industry-wide: Mistral's new Large 4 is also being pre-released to cybersecurity experts and state authorities with fewer safety barriers before its October 27 public launch.


AiPost's take: The cybersecurity game has flipped. AI is now both the fastest bug hunter and the tool everyone is nervous about handing to the wrong people. Defense-first access looks like the sanest way to run the experiment.

Comments

Post a Comment

Popular posts from this blog

Top 5 Best AI Tools in 2026 (I Tested Them All)

-
Image
Look, I'll be honest with you. Every other day someone asks me "which AI should I actually use?" — and every other day there's a new tool claiming to be the best. So I did what any curious person would do: I spent weeks actually using these tools, not just reading about them. Here are the 5 AI tools that genuinely earned a place in my daily routine in 2026. No hype, no sponsored picks — just what actually works. 1. ChatGPT Best for: Almost everything — the safest all-rounder If you only pick one AI tool this year, make it this one. I've used ChatGPT for writing emails, debugging code, planning trips, learning new skills — you name it. It's fast, it remembers context well, and the ecosystem around it (voice mode, image generation, custom GPTs) is unmatched. It's like that friend who's decent at everything. Price: Free plan available; Plus at $20/month unlocks the best models. Try ChatGPT → 2. Claude Best for: Writing, long document...
Read more

SpaceXAI's Grok 4.7 Is a Bigger, Smarter Model at the Same Price

-
Image
This week everyone in AI was slashing prices. SpaceXAI, Elon Musk's AI company, took a different road: on September 21, 2026 it released Grok 4.7 , a bigger, smarter model — at exactly the same price as its predecessor. The company calls it its most capable model yet for coding and knowledge work, built on a new, larger base model than Grok 4.6 and trained with a longer reinforcement learning run weighted toward tasks that take hours to finish rather than seconds. The stated goal: a model that sticks with hard problems longer, double-checks its own output, and handles long context windows without losing the thread. What's new under the hood Four changes separate Grok 4.7 from the last version: the new base model itself, a longer and harder RL training run, upgraded self-verification and long-context handling, and native training on the Grok Bot harness, which SpaceXAI says improves conversational tasks and general knowledge work — documents and presentations inclu...
Read more

YouTube Just Rolled Out a Batch of AI Tools for Creators — Here's What They Do

-
Image
YouTube wants AI to sit beside every creator — from the first script idea to the thumbnail that finally gets the click. At its annual Made on YouTube event in New York City on Wednesday, the company unveiled a slate of AI tools for creators, covered by Reuters and TechCrunch the same day. What's new The biggest changes land in YouTube Studio , the app creators use to run their channels. The standout is AI feedback on drafts : upload an unpublished video and the tool critiques its pacing, structure, and storytelling before it ever goes public. Alongside it comes a new research feed that shows what's working on the platform right now, so creators can put their own spin on it. Thumbnails — the single highest-leverage pixel on YouTube — get the deepest upgrade. Studio can now generate thumbnails and titles from a video's content, matched to the creator's style . A new dynamic thumbnail feature creates three options and shows each one to a different slice of the audi...
Read more