|
|
|
|
|
|
|
|
|
|
OpenAI to Hack Its Own Models to Identify 'Catastrophic' Risk in AI
|
|
|
|
|
|
|
|
Hello. ChatGPT maker OpenAI is putting together a team to assess the capabilities of—and risks created by—frontier-pushing uses of AI in several areas, including cybersecurity and chemical and nuclear threats.
Aleksander Madry, a computing professor on leave from the Massachusetts Institute of Technology, will lead the so-called preparedness team, OpenAI said Thursday.
Also on the team will be red-team hackers whose job will be to break frontier AI systems so they can be fortified against cyberattacks and malicious misuse, the company said. According to a job post, OpenAI is seeking artificial-intelligence experts who can identify "misuse risks whose impact could be catastrophic (not necessarily existential) to our society."
The move comes as President Biden prepares to issue an executive order on "trustworthy AI" and the U.K. hosts the AI Safety Summit on Wednesday and Thursday.
Also today:
-
Seiko data leaked in cyberattack
-
Security overhaul at twice-breached LastPass
-
AI security startup Cranium raises $25 million
|
|
|
|
|
|
|
|
|
CONTENT FROM: Netscout
|
|
Top 6 DDoS Lessons from 1H 2023
DDoS attacks continue to pose detrimental threats to networks across the globe. NETSCOUT’s most recent threat report provides an in-depth analysis on DDoS infrastructure, attack patterns, and the best defense strategies. Explore the six key highlights from our report and stay ahead of cyber threats.
Learn More
|
|
|
|
|
|
|
|
|
|
|
|
PHOTO: ARND WIEGMANN/REUTERS
|
|
|
|
|
Seiko discloses compromise of personal data of employees, customers and business partners. A ransomware attack detected in late July, and revealed by the Tokyo-based company in August, exposed the information. Seiko Group, Seiko Watch and Seiko Instruments were affected. The company continues to investigate, it said Wednesday.
|
|
|
|
Canadian man pleads guilty in a series of cyberattacks. A 33-year-old man targeted more than 1,000 individuals and mainly small organizations in Canada and the U.S., according to prosecutors. Through ransomware and access to victim bank accounts, he stole about $49,000, Ontario police said. He hasn't yet been sentenced. (Ottawa Citizen)
|
|
|
Breached tech provider LastPass is overhauling its cybersecurity. A multi-million-dollar investment in new security tools and systems is underway at LastPass, said Karim Toubba, chief executive of the password manager company. Hackers compromised LastPass twice in 2022, accessing business and customer information. The company saw customer churn of 9% in late 2022, Toubba said, adding that churn has improved and he expects to see pre-incident levels next year. (Cybersecurity Dive)
|
|
|
AI security startup Cranium said it has raised $25 million in a Series A funding round led by Telstra Ventures with KPMG LLP and SYN Ventures. Cranium, based in Short Hills, N.J., is working on tools to gather information about the trustworthiness and compliance of customers' AI models. It was spun out of KPMG's tech incubator.
|
|
|
|
|
Our weekly roundup of stories from across WSJ Pro that we think you'll find useful. They are unlocked for WSJ subscribers.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
