Security — page 18
OpenAI publishes Frontier Governance Framework to align with EU and California rules
The document maps OpenAI's safety, security, and risk practices to emerging regulations in the EU and California.
Illinois passes frontier AI safety bill requiring third-party audits
SB 315 mandates independent verification of safety practices at OpenAI, Anthropic, and Google DeepMind.
BadHost vulnerability exposes millions of AI agents to credential theft
CVE-2026-48710 affects Starlette, a Python package with 325 million weekly downloads, allowing trivial authentication bypass.
Universal Music Group and TikTok renew licensing deal to combat unauthorized AI music
The renewed agreement commits both companies to removing AI-generated tracks and improving artist attribution on the platform.
AI warfare is already here as Anthropic fights Pentagon over autonomous weapons
The DOD has embedded AI in military operations for 70 years; fully autonomous lethal systems remain the final frontier.
AI bug hunting arms race doubles vulnerability payouts as exploit discovery accelerates
Google and Apple face 2–10x more bug submissions; researchers report tripled output, forcing bounty programs to rebalance economics.
Google Cloud's AI security gap: deleted API keys work for 23 minutes
Compromised Gemini keys can authenticate for nearly half an hour after deletion, exposing files and cached data.
Ocean raises $28M to fight AI-powered phishing with agentic email defense
The startup's small language model analyzes billions of emails monthly for customers including Kayak and Kingston Technology.
Former OpenAI staffers warn xAI safety record could complicate SpaceX IPO
A new watchdog group says xAI's safety incidents are disproportionate to its market share and should trigger investor disclosures.
Elon Musk loses OpenAI lawsuit on statute of limitations ruling
A jury found Musk should have sued by 2021, three years before he filed — the case never reached the merits.
Jury dismisses Elon Musk's lawsuit against OpenAI and Sam Altman as untimely
The verdict ends a 3-week trial over Musk's claims that Altman betrayed OpenAI's nonprofit mission.
Bug bounty programs buckle as AI-generated slop reports flood inboxes
Bugcrowd saw submissions quadruple in three weeks; Curl and Nextcloud suspended their programs as low-quality AI reports pile up.
Anthropic to brief global financial watchdog on Mythos cyber flaws
The Claude maker is set to present findings on vulnerabilities exposed by the Mythos incident to international financial regulators.
ArXiv will ban authors for a year if LLMs do the work unchecked
The preprint repository's computer science chair Thomas Dietterich set a one-strike rule for papers showing hallucinated citations or stray LLM commentary.
CFTC turns to AI surveillance to police insider trading on Polymarket and Kalshi
Chairman Michael Selig says the agency is chasing 'hundreds, if not thousands' of tips with AI-assisted pattern detection and blockchain tracing.
Musk v. Altman trial closes with $134B at stake and an OpenAI IPO in the balance
Closing arguments wrapped in Oakland as Musk seeks to unwind OpenAI's 2025 restructuring; the jury begins deliberating Monday.
YouTube opens AI likeness detection to all users 18 and older
The facial-match deepfake scanner, previously limited to creators, officials, and entertainers, now covers any adult with a YouTube account.
Authors object to $320M legal fees in Anthropic's $1.5B copyright settlement
A federal judge delayed final approval after class members called the $3,000-per-author payout a 'pittance' next to lawyer fees.
ArXiv will ban authors for one year over unchecked AI-generated paper content
The preprint server's computer science section chair, Thomas Dietterich, says hallucinated references and leftover LLM meta-comments now trigger a 12-month suspension.
arXiv will issue one-year bans for AI-generated slop in preprints
Authors caught submitting AI-generated errors, fake citations, or misleading content face a 12-month ban and a permanent peer-review requirement.
US judge weighs Anthropic's $1.5B settlement with authors over book training data
The proposed deal would resolve a class action alleging Anthropic trained Claude on pirated books, pending court approval.
Nine California jurors begin deliberations in Musk v. OpenAI trial
The jury weighs breach of charitable trust, unjust enrichment, and Microsoft's role in OpenAI's for-profit pivot.
Princeton ends 133-year no-proctor tradition as 30% of seniors admit AI cheating
Faculty voted to require instructor proctoring from July 1 after a 2025 survey found 29.9% of seniors cheated and 44.6% witnessed it without reporting.
xAI runs 46 gas turbines at Mississippi data center under 'mobile' loophole
Mississippi treats the trailer-mounted turbines as mobile, exempting them from air-pollution rules for one year; the NAACP is seeking an injunction.
Stay ahead of everyone in AI.
The tightly edited AI news email engineers, founders, and investors actually open. One email. Every weekday. Five minutes to finish.
The briefing read inside teams at