AI agents breached Anthropic systems in bug bounty test
The Decoder: AI agents hacked Anthropic, maker of arguably the most capable AI models for cyberdefense
Cybersecurity startup Cathedral used AI agents to break into Anthropic's internal systems through third-party software as part of a bug bounty program. Anthropic confirmed access to sales, HR, and IT data but disputes how deep the breach went. The incident shows how vulnerable even leading AI labs are through their external software dependencies. The article AI agents hacked Anthropic, maker of arguably the most capable AI models for cyberdefense appeared first on The Decoder.
Get smart on it
A cybersecurity startup used AI agents to break into an AI company's internal systems through third-party business software as part of a bug bounty program, accessing internal data and messaging systems. The incident matters because it demonstrates that AI agents can now probe corporate networks continuously at minimal cost, and it occurred at a company widely considered among the most capable in AI-based cyberdefense. The breach exposed disagreements about scope and duration, with disputes over whether source code and model weights were compromised, and raised questions about whether even the most advanced AI companies can adequately secure their systems against AI-powered attacks. Security experts cited in the reporting indicated that integrated third-party applications represent a widespread industry vulnerability and that the pace of AI development may be prioritizing speed over security measures.
Relevant courses
Want to go deeper than the news? Explore live, cohort-based AI courses taught by practitioners.
Browse AI courses on MavenRelated news
Parents sue OpenAI, alleging GPT-4o coached their daughter off seizure medication
A family is suing OpenAI, saying ChatGPT coached their daughter to quietly taper off her seizure medication and keep it from her doctors before she died in February.
Alibaba's latest Qwen models routed some requests through company servers, researchers find
Alibaba's Qwen 3.8 models tried to send data to the company's own cloud servers during testing. Palisade Research finds the behavior in three models from the 3.8 generation, while older models aren't affected. The likely cause: Alibaba's training infrastructure got baked into the models as a learned behavior. The article Alibaba's latest Qwen models routed some requests through company servers, researchers find appeared first on The Decoder.
Google AI gets cagey when asked who won the 2020 elections
If you ask some prominent US politicians who won the 2020 presidential election, you might get some weasely answers that fail to directly acknowledge Joe Biden's win that November. If you ask Google's AI Mode the same question, you might get an evasive answer that simply tells you to consult with outside sources. When I put the direct question of "Who won the 2020 presidential election?" to Google's AI Mode this morning (after seeing complaints about cagey answers on social media), I got the fol
Meta says it’ll probably delay its controversial new business terms again
Meta says it may push the implementation of its new business terms past November 20th as it faces backlash from publishers over concerns about AI training. In a post on X, Meta communications head Andy Stone said the company does "not plan to make policy changes without a full consultation with industry groups and leaders." The new business terms, which were supposed to take effect on October 30th, initially appeared to give Meta permission to train its AI models on the news si
OpenAI's isolated AI agents found ways to talk to each other through CPU cache
OpenAI now trains its most powerful models without internet access, after an internal research model exploited a DNS misconfiguration in September and reached the open internet. In controlled tests, AI agents still found ways to communicate covertly through side channels. The incidents echo Sam Altman's warning: the more powerful the model, the harder it is to contain. The article OpenAI's isolated AI agents found ways to talk to each other through CPU cache appeared first on The D
Three-week on-site audit at OpenAI reveals major gaps in AI safety monitoring
External auditors from METR and Redwood Research spent three weeks on-site at OpenAI examining the company's internal monitoring system for AI agents and found significant gaps, with up to 25 percent of agent traffic from models at the GPT-5.5 class and above going unmonitored. OpenAI redacted large portions of the 61-page report, and only a summary is publicly available. The article Three-week on-site audit at OpenAI reveals major gaps in AI safety monitoring appeared first on The Deco