AI Summary
5 min readAnthropic’s bug-hunting AI, Mythos, found 90 critical and 141 important bugs in Microsoft SharePoint in April alone—and it’s finding them faster than Microsoft’s engineers can patch. Microsoft shipped 600 fixes on July 14, and internal meetings show the company was racing a May 31 deadline before hostile governments build equivalent tools. Since the start of the year, Mythos has also found hundreds of critical bugs in Microsoft 365, Teams, and Copilot. About 300 moderate-severity SharePoint flaws are still queued for patching. One of the bugs Mythos found was already being exploited by hackers in the wild. The Five Eyes Intelligence Alliance warned in late June that the window for defenders to outpace AI-equipped attackers would close within months; internal Microsoft documents suggest it may have already closed.
The real danger is that Mythos can chain multiple small bugs into working exploits that humans miss. Vin Newgen, a senior Anthropic advisor and former NSA AI chief, said Microsoft’s standard triage system—patch critical flaws first, low severity last—no longer works when AI can weaponize the bottom of the queue. Speed alone might not be enough.
Google’s Synth ID Watermark Survives 300 Edits—But Fragmentation Looms
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Hard Fork AI
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:00) **Anthropic's Mythos AI Finds Critical Bugs Faster Than Microsoft Can Patch** - A bug-hunting AI discovered 90 critical and 141 important bugs in Microsoft SharePoint in April, outpacing Microsoft's engineers.
- 2 (03:25) **Google's SynthID Watermark Survives 300 Edits, But Fragmentation Remains** - An invisible watermark on AI-generated images survived extensive compression and resizing, but its utility is limited by platform fragmentation.
- 3 (07:17) **Encore AI Raises $30 Million to Build Voice Agents Trained on Real Sales Calls** - The startup analyzes transcripts of successful sales calls to train AI agents on what top performers say to close deals.
- 4 (10:49) **Writers Are Deliberately Breaking Style Rules to Fight AI Homogenization** - Writers and editors are adding typos, using all lowercase, and avoiding em dashes to make their work unmistakably human.
- 5 (13:37) **Sierra Acquires Oasis Security for $1 Billion to Manage AI Agent Identities** - The $12 billion data security company is buying Oasis to track and manage the non-human identities of AI agents inside corporate systems.
- 6 (15:46) **Google's DeepMind Ships Lyria 3.5 with Better Vocals and Lyrics** - The updated music generation model improves on vocals and lyrics, challenging competitors like Suno and Udio.
- 7 (16:26) **Panagram Raises $9 Million to Detect AI-Generated Text with 99% Accuracy** - A new AI text detector claims high accuracy, but the host is skeptical given the cat-and-mouse nature of detection.
+ Full timestamped outline available in the app
Show Notes
Chapters
00:00 Microsoft Bug Hunting AI
02:01 Google's SynthID Watermark
04:59 Encore AI's Voice Agents
10:16 Writers Embrace Typos
14:59 Sierra Acquires Oasis Security
17:01 Recent Industry Headlines
Get the top 80+ AI Models for $8.99 at AI Box: https://aibox.ai
How I Grow and Scale My Business with AI: https://www.skool.com/aihustle
Get the AI Chat Daily Newsletter: https://www.aichatdaily.com/newsletter
More from this podcast
Hard Fork AI →