Information Security Briefly@infosecbriefly.bsky.social
Daybreak expands access to GPT-5.6-Cyber for authorized defensive security work via Blue and Red routes with different guardrails and completion rates.
I have been thinking about this for a while now.
"% of secrets in the vaults" feels like it should be a table-stakes number for compliance and governance programs. blog.gitguardian.com/vault-covera...
“If a rogue AI model broke out of its security infrastructure for days without detection, accessed the internet despite precautions against such connection, and hacked into other firms in defiance of human commands, this could have serious implications for America’s national security.”
CISA, the FBI, and international partners have issued a warning about Gunra ransomware-as-a-service operation, which is targeting critical infrastructure entities, including healthcare organizations. zurl.co/99jky
Qilin ransomware hit Tommer Construction in the US, encrypting critical files and disrupting operations as the group demanded ransom for data restoration. #UnitedStates#TommerConstruction#Ransomware
Flight tracking platform FlightAware sues Kalshi in New York, alleging Kalshi is using its data without permission to let users bet on flight cancellations (Jack Morphet/Wall Street Journal)
Brad Lightcap, who most recently led OpenAI's special projects division and formerly was its COO, says he is leaving the company "to start something new" (Stephanie Palazzolo/The Information)
A24's $75 million partnership with Google DeepMind has indie movie lovers buzzing—some are outraged! As AI's grip on Hollywood tightens, the debate intensifies. Is this innovation or intrusion? What do you think? #A24#AIi…
Scammers don’t always need sophisticated technology. Sometimes, the victim is the exploit. I got bogus text last week & ignored it. Others weren’t so lucky. In less than 4 minutes, scammers convinced victims to hand over their debit-card numbers & sensitive infor. brothke.medium.com/the-victim-i...
Qilin ransomware targeted Service Evaluation Concepts, a US professional services firm, encrypting or disrupting systems and demanding a ransom. #UnitedStates#Ransomware#ProfessionalServices
OpenAI debuts GPT-5.6-Cyber, a security model that answers 98.5 percent of previously blocked queries and has already found two unknown Chrome vulnerabilities. Access requires identity verification.
. @cmu.edu faculty and students will present 10 co-authored papers on a wide range of topics at the 35th USENIX Security Symposium, held this week in Baltimore. See a full list of papers and abstracts below: #usesec26
NEW: Researchers found vulnerabilities in Zoom that let an attacker take over a device on any call with screen sharing. They used publicly available AI models and fewer than 20 prompts. Zoom has since patched. The AI-led democratization of 0 days continues apace.
Exploitation is getting faster, Content Management Systems remain the top target, and AI is emerging as a new attack surface despite low observed exploitation rates.
AI is your newest teammate—a hyper-productive contributor that requires strong guardrails to succeed. Read more on why this partnership makes your choice of programming language more important than ever. developers.googleblog.com/why-go-is-an...
if I had fully forgotten how to read, you would not be able to torture it out of me. but this guy wrote a whole article about it! www.axios.com/2026/08/11/r...
Signal or noise? > An unreleased Anthropic model made progress on one of math’s
biggest unsolved problems >> Comment below! #AI#industry40#IoT#mhealth#healthtech
American University School of International Service@ausis.bsky.social
Professor @willakoto.bsky.social explains how and why hackers are targeting water systems and what utility companies can do in defense in this article originally published by @us.theconversation.com.
OpenAI first said these were “results to problems that have been open and have seen no progress on the main result for at least a decade, and in most cases much longer.” It later changed it to “results, each of which resolves or makes substantial progress on a long-standing open problem.”
Eight researchers left OpenAI with Dario Amodei in 2021 to build Anthropic. Now Anthropic's own alumni are raising $680M for their own AI labs. The safety diaspora keeps spreading:
