{"api_version": 1, "episode_id": "ep_hard_fork_786bd62e8eb9", "title": "Anthropic\u2019s Cybersecurity Shock Wave + Ronan Farrow and Andrew Marantz on Their Sam Altman Investigation + One Good Thing", "podcast": "Hard Fork", "podcast_slug": "hard_fork", "category": "tech", "publish_date": "2026-04-10T11:00:00+00:00", "audio_url": "https://dts.podtrac.com/redirect.mp3/pdst.fm/e/pfx.vpixl.com/6qj4J/pscrb.fm/rss/p/nyt.simplecastaudio.com/3e43d072-f8a5-430f-bc8e-4c70aafdf3c7/episodes/29726999-09f3-4807-964d-929b2d55994c/audio/128/default.mp3?aid=rss_feed&awCollectionId=3e43d072-f8a5-430f-bc8e-4c70aafdf3c7&awEpisodeId=29726999-09f3-4807-964d-929b2d55994c&feed=l2i9YnTd", "source_link": "https://www.nytimes.com/column/hard-fork", "cover_image_url": "https://image.simplecastcdn.com/images/4105a47a-42e5-4ccc-887a-832af7989986/23965394-f5e4-4fdb-b150-639f4910353e/3000x3000/nyt-hf-album-art-3000-2.jpg?aid=rss_feed", "summary": "Anthropic announced Project Glasswing and its Claude Mythos preview model, which reportedly found critical zero-day vulnerabilities in OpenBSD and FFmpeg, prompting a consortium of major tech firms to conduct defensive cybersecurity testing. The model is not being released publicly due to its potential for misuse, with Anthropic arguing that proactive hardening of infrastructure is necessary. Experts like Alex Stamos suggest this could trigger a massive global software patching effort within six months.", "key_takeaways": ["Claude Mythos preview found a 27-year-old security flaw in OpenBSD and a previously undetected bug in FFmpeg despite 5 million prior scans.", "Anthropic is restricting access to the model, sharing it only with a consortium including Microsoft, Apple, and Amazon for defensive cybersecurity use.", "The rollout reflects a shift toward AI-driven, autonomous discovery of zero-day exploits at speeds far beyond human teams."], "best_for": ["AI and cybersecurity professionals", "tech policy analysts", "software engineers managing open-source dependencies"], "why_listen": "You\u2019ll understand how AI is shifting from tool to autonomous threat actor in cybersecurity, with real examples of exploits found in foundational software.", "verdict": "must_listen", "guests": [], "entities": {}, "quotes": [], "chapters": [], "overall_score": 90.0, "score_breakdown": {"clarity": 90.0, "originality": 92.0, "actionability": 75.0, "technical_depth": 88.0, "recency_relevance": 100.0, "information_density": 95.0}, "score_evidence": {"clarity": "the intelligence in these machines and their ability to work autonomously are now great enough that they can chain together exploits", "originality": "they're not releasing this model to the public because they claim it is too dangerous to do that", "actionability": "they can basically use this model to proactively go out and find all of the unfound bugs, they call these zero day exploits", "technical_depth": "this model has found vulnerabilities in every major operating system and web browser", "recency_relevance": "On Tuesday, Anthropic announced that it was starting something called project glasswing", "information_density": "found a twenty seven year old security flaw in OpenBSD... found a bug in FFmpeg that had been scanned 5,000,000 times without detection"}, "score_reasoning": {}, "scoring_confidence": 0.95, "transcript_available": true, "transcript_chars": 67968, "transcript_provider": "deepgram"}