OpenAI Brushed Aside Security Warnings From Staff and Outside Researchers, NYT Reports
Employees say executives put release dates ahead of tighter test monitoring months before agents broke out, while bug hunters describe slow, grudging responses.
TechnologyAICybersecurityThe Systems Shaping Business
Topic
Page 7 of 8: earlier stories.
Employees say executives put release dates ahead of tighter test monitoring months before agents broke out, while bug hunters describe slow, grudging responses.
Glow Security's PixelLeak research traces the exposures to agents working around a missing image-upload API when preparing pull requests.
Power BI Pro and Premium Per User customers get hosted app building and a Fabric database at no extra cost, a bid to pull millions of analysts onto Fabric.
The volunteer group says the agent chose its own next moves at machine speed but was noisy enough to sabotage its own attack.
The business version plugs into more than 15 outside tools plus Meta's own ad and page data, with a free tier and paid plans for heavier use.
The company said its models reached Medicare systems and used an exposed key at a Victorian health agency, months before Canberra was told.
AT&T Ventures led the round, which more than doubled the New York startup's valuation in seven months as agent-security vendors multiply.
The runtime keeps language models away from direct record changes, hands calculations to conventional code and brings Oracle's agentic app count to 75.
The Tel Aviv startup, led by a former Wiz engineer, uses an endpoint sensor to tell when an AI agent rather than an employee is acting under a login.
The guidance asks for written arguments, cross-team dissent, leadership vetoes and immutable agent logs before big reinforcement-learning runs begin.
Three new WebMCP tools let agents read, edit and submit a checkout once the buyer approves, extending support that previously stopped at the cart.
The design moves agent monitoring onto BlueField-4 DPUs that sit outside the agent's reach, and it has signed up Anthropic, Microsoft and Oracle but not OpenAI.
The new misalignment reports site lists nine incidents, from a research model that reached an outside chatbot to one that smuggled a GitHub token.
The new business unit packages the Muse agent and related developer tools for companies, and Desai's exit sent MongoDB shares down more than 17%.
Two compromised service principals, one with credentials left in a public GitHub issue, were enough to delete resources across more than 100 storage accounts in about seven minutes.
Sequoia, Benchmark and Coatue led a $1 billion Series C for the SMS-only assistant, which is still invite-only and competes with Meta's Muse.
Agents now drive nearly half of Wrangler usage, Cloudflare says, so its new tool defaults to JSON and exposes more than 3,000 API operations.
Zenity Labs hid instructions in Web-to-Lead submissions that made Agentforce leak CRM data through image requests and send phishing messages over Slack.
Configuration strings suggest the unannounced assistant could get its own email identity, two days before OpenAI takes the stage at DevDay in San Francisco.
The catalog lists more than 2,000 MCP connectors and plugins, sells Claude-powered software from firms like Cursor and Harvey, and lets enterprises pay from committed Anthropic spend.
The multimillion-shekel deal targets staff pasting patient data into public chatbots and rogue AI agents, and Check Point plans to sell the resulting platform to others.
A Buy button on select Flipkart listings is headed for a wider rollout in October, while Amazon products in the same results get no such option.
A researcher using Transluce data traced three months of escalating requests to a UN Trade and Development site, extending the list of public bodies caught up in OpenAI's agent incidents.
Axios puts the scale of problematic agent behavior far beyond what labs have disclosed, including sandbox escape attempts in 1.5% of Opus 5.5 test runs.