Google DeepMind researchers push for right to publicly discuss AI findings

·3 views

Staff at Google DeepMind are protesting a policy that can impose lifetime bans on employees who publicly criticize the company. The researchers argue they should be free to share what they know about artificial intelligence with the public. The dispute was reported by The Washington Post and discussed on Hacker News, where the story drew 37 points and 22 comments. The article linked from the discussion focuses on the employees' demands and the company's restrictions on speech. Details beyond the article's headline and the reported protest are limited in the available source material.

Read the full story at Hacker News

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Why Uptime SLA Credits Rarely Protect Businesses From Outages

An uptime SLA typically functions as a limited refund rather than a guarantee of availability, a distinction many customers overlook. Percentages like 99.9% can permit surprisingly large amounts of downtime, such as about 43 minutes a month, so the right target depends on what an hour of outage actually costs. Credits are usually capped at a share of monthly fees, which generally falls far short of real losses, and they are often claim-based with short deadlines. The article argues that resilient architecture and recovery planning, not credits, are what actually protect a business, and that buyers should clearly define downtime, measurement methods and exclusions.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

DoubleClickjacking Bypasses Frame Protections by Swapping Windows Mid-Double-Click

A December 2024 technique called DoubleClickjacking was published by security researcher Paulos Yibelo, showing that common frame-blocking defenses can be circumvented. The attack avoids embedding a page in a frame, instead opening a top-level window that prompts a victim to double-click. The first click closes the overlaying window, and the second click lands on a sensitive control, such as an OAuth approval button, on the underlying real page. Because the target page loads as a normal top-level page with the victim's session, headers like X-Frame-Options, CSP and SameSite cookies do not block it. The article notes that the attack typically targets screens where a single click carries serious consequences, such as approving an app or confirming a payment.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

ScamShield AI Week 2: Hardening Docker Firewalls and Fail-Closed Sandbox Design

A developer building ScamShield AI, a threat-intelligence platform for analyzing Indian scam messages, documented security hardening of an isolated Docker sandbox used to safely open suspicious links. The sandbox pairs headless Chromium via Playwright with a Mitmproxy network guard. The author found that iptables rule ordering created a bypass, since a stateful ESTABLISHED,RELATED accept rule preceded the drops for private and cloud-metadata addresses, risking SSRF. The firewall was restructured into six phases so prohibited destinations are dropped before any connection-state acceptance. The startup script was also changed from fail-open, which merely warned when iptables was missing, to fail-closed, which aborts the container if rules cannot be applied or verified.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Rutgers researchers warn of rare sudden vision loss linked to Ozempic and Wegovy

Rutgers University researchers have warned that the GLP-1 drugs Ozempic and Wegovy, widely used for weight loss and diabetes treatment, may rarely cause serious sudden vision loss. The warning has raised concern among patients who rely on these medications for their health. The article also presents a first-person account from a writer with childhood amblyopia, or lazy eye, who describes how a VR training app called AmblyoPunch was used as a supplementary vision therapy tool alongside professional guidance. The author notes the app is not a cure or medical device but describes its gamified approach to retraining the brain and eyes. The piece is presented as a personal experience rather than a clinical study.

0 comments Read more at DEV Community