Tag

Jailbreak

All articles tagged with #jailbreak

technology10 days ago•4 min saved

Anthropic's Claude Fable 5 Returns Online as US Lifts Export Rules

The U.S. Commerce Department has lifted export controls on Anthropic’s Claude Fable 5, allowing the model to go back online worldwide on July 1 across Claude.ai, the Claude Platform, Claude Code, and Claude Cowork. The controls had blocked access for foreign nationals after a jailbreak prompt could bypass safety rules; Anthropic says it patched the issue with a new safety classifier that blocks the technique in over 99% of attempts and reroutes problematic requests to the weaker Opus 4.8. Mythos 5 remains under tighter restrictions. Access began for about 100 U.S. companies and federal agencies on June 26, with plans to widen further; the company is introducing a standardized jailbreak severity scoring system and a HackerOne program for researchers, reflecting ongoing governance and risk-management efforts around frontier AI models.

via The Hacker News|

#ai-safety #anthropic #claude-fable-5

technology23 days ago•5 min saved

US moves toward AI security standards with Anthropic as export controls linger

The White House and Anthropic are crafting a framework to benchmark security flaws in new AI models and guide potential government action, signaling a shift from export controls to a standardized, standards-setting process as talks continue.

via Politico|

#ai-security #anthropic #export-controls

technology25 days ago•2 min saved

Anthropic dispute unsettles US AI security research and cyber defenses

A feud over access to Anthropic's Fable 5 and Mythos 5 has security leaders warning that punitive moves could chill defensive AI research and weaken U.S. cyber defenses, as the administration weighs new AI security rules and a forthcoming vulnerability-clearinghouse, with critics arguing the conflict risks advantaging adversaries and undermining defenders' capabilities.

via Axios|

#ai-security #anthropic #cybersecurity-policy

technology26 days ago•7 min saved

Anthropic, White House Stay Complicated Over Claude Fable 5 and Jailbreak Concerns

The White House has not lifted export controls on Anthropic’s Claude Fable 5 after Monday talks, with officials worried the jailbreak risks could enable Mythos-like capabilities; Anthropic argues safeguards are strong and presses for a path to consumer use, as Commerce Department, NSA, and company leaders continue negotiations that will influence how advanced AI models are released and regulated in the U.S.

via WIRED|

#ai-policy #anthropic #claude-fable-5

technology26 days ago•4 min saved

Policy clashes push Anthropic's flagship AI offline

Anthropic's Mythos and Fable were taken offline after the White House imposed export controls tied to a cyber executive order; Anthropic says it had explicit government approval to deploy Fable, while officials say the company failed to heed the order and distributed a jailbreak. The spat highlights regulatory and communication hurdles as the government moves to govern frontier AI, with Commerce, CIA, and White House science staff set to meet to align on safeguards and compliance.

via Axios|

#anthropic #export-controls #fable

technology29 days ago•10 min saved

Anthropic halts Fable 5 and Mythos 5 under export-control directive

Anthropic disabled access to its Fable 5 and Mythos 5 models after a US government export-control directive to restrict their use outside the US, citing a narrow potential jailbreak and ongoing national-security considerations; other Anthropic models remain available and the company plans to provide more details soon.

via Ars Technica|

#anthropic #export-controls #fable-5

technology2 months ago•11 min saved

AI jailbreakers push safety to the edge by coaxing dangerous outputs from chatbots

A growing community of ‘jailbreakers’ tests large language models by manipulating prompts and social tactics to bypass safety rules, revealing how even frontier AI systems can be coaxed into dangerous outputs. The piece profiles practitioners like Valen Tagliabue and David McCarthy, explains how firms patch vulnerabilities, and underscores the ongoing risk as AI becomes more capable and integrated into everyday devices and workflows.

via The Guardian|

#ai-safety #ethics #frontier-models

technology2 months ago•8 min saved

Kindle Obsolescence: Jailbreaks and the DRM-Free Reading Route

As Amazon ends support for older Kindle models, readers are urged to jailbreak the device to install alternative e-reading apps, transfer content via Calibre, and switch to DRM-free ebooks from sources like Project Gutenberg and Cory Doctorow, arguing that buyers own their content and can read it on non-Amazon devices.

via Boing Boing|

#amazon #calibre #drm

technology6 months ago•2 min saved

Star Wars Game Sparks PS5 Jailbreak and High eBay Prices

A rare physical PS4 disc of Star Wars Racer Revenge has become highly sought after due to its role in a new PlayStation 5 jailbreak exploit targeting firmware 12.00, causing its price to skyrocket from $20 to nearly $400, driven by its unique unpatched vulnerability and limited availability.

via TechSpot|

#collectors-market #console-hacking #jailbreak

crime6 months ago•2 min saved

Georgia Jail Escapees Hijack Lyft for Florida Getaway

Three men, all convicted of violent crimes including murder, escaped a Georgia jail and hijacked a Lyft to flee to Florida, where they used the driver’s credit cards and vehicle before being recaptured. The incident involved threats and violence against the driver, and the men were tracked down using license plate readers after a multi-state pursuit.

via The Guardian|

#crime #escape #florida

crime6 months ago•1 min saved

Three inmates, including murder suspect, escape Georgia jail

Three inmates, including a murder suspect, escaped from DeKalb County Jail east of Atlanta, prompting a serious search by law enforcement; the escape's method is unknown, and the inmates are considered armed and dangerous.

via ABC News|

#crime #dekalb-county #escape

technology9 months ago•8 min saved

ChatGPT safety flaws enable weapon instruction bypass

OpenAI's ChatGPT guardrails can be bypassed using jailbreak prompts, allowing some models to generate dangerous instructions for weapons and chemical agents, raising concerns about AI safety and misuse, especially as open-source models are more vulnerable and could be exploited by bad actors.

via NBC News|

#ai-vulnerabilities #bioweapons #chatgpt

crime9 months ago•6 min saved

Final New Orleans jailbreak escapee captured in Atlanta, waives extradition

Derrick Groves, the last of ten inmates who escaped from a New Orleans jail in May, was captured in Atlanta after five months on the run, ending a multi-agency manhunt. Groves, convicted of murder and other charges, is expected to be extradited to Louisiana, where he faces additional charges and potential upgrades to his escape-related charges. The jailbreak involved systemic issues and multiple arrests of accomplices, highlighting ongoing concerns about prison security.

via NPR|

#arrest #crime #escapee

crime9 months ago•5 min saved

Final New Orleans Jail Escapee Captured After Five-Month Manhunt in Atlanta

Derrick Groves, the last of 10 inmates to escape from a New Orleans jail in May, was captured in Atlanta after a months-long search, ending one of the largest recent U.S. jailbreaks. He was found hiding in a crawl space and is now facing extradition and additional charges.

via AP News|

#atlanta #crime #derrick-groves

crime9 months ago•3 min saved

Final New Orleans Escapee Derrick Groves Captured in Atlanta

Derrick Groves, the last of 10 inmates who escaped from a New Orleans jail nearly five months ago, was captured in Atlanta after a standoff, ending a lengthy manhunt. Groves, convicted of double murder, was apprehended following a tip and a multi-agency effort, bringing relief to victims and authorities.

via CNN|

#arrest #crime #derrick-groves