All news with #openai api tag

🔒 OpenAI announced it has temporarily slowed scaling, paused frontier reinforcement learning runs, and will offer zero data retention for eligible API customers starting in September. The company said it hardened its research environment, expanded monitoring, and will require stronger evidence of aligned behavior during training. Analysts say the moves may be aimed at shoring up trust before an IPO, while critics call some steps theatrical without regulatory or contractual commitments.

🤖 OpenAI has reduced API prices for two GPT-5.6 models, cutting Luna by 80% and Terra by 20% to improve cost efficiency. Luna now costs $0.20 per million input tokens and $1.20 per million output tokens, down from $1 and $6; Terra’s rates dropped to $2 per million input and $12 per million output. OpenAI also updated usage accounting for Codex and ChatGPT Work and upgraded Auto-review to GPT-5.6 Luna, yielding significant cost savings. Additionally, GPT-5.6 Sol gains a Fast API option that is up to 2.5× faster at twice the price for latency‑sensitive workloads.

🛡️ Sophos analyzed a week of June 2026 telemetry and found AI coding agents like Claude Code, Cursor, and OpenAI Codex frequently trigger behavioral detection rules designed to catch human attackers. The agents perform actions—decrypting browser credentials, enumerating Windows Credential Manager, downloading files via LOLBins, and writing startup scripts—that look like malicious behavior to endpoint engines. While often benign developer automation, these behaviors overlap precisely with attacker techniques and can generate false positives. Sophos recommends scoping rules to agent parents, workspaces, and download reputations while keeping credential access tightly controlled.

🖥️ Amazon Bedrock AgentCore Runtime introduces the InvokeAgentRuntimeCommandShell API, providing a persistent, PTY-backed terminal over WebSocket into running agent sessions. This complements existing one-shot execution via InvokeAgentRuntimeCommand and delivers a full terminal experience inside an isolated microVM with features like colors, tab completion, Ctrl+C, resize, and automatic reconnect. Developers hosting coding agents (for example, Claude Code, OpenAI Codex, Amazon Kiro) can now authenticate, drop into the agent microVM, inspect files, run ad-hoc commands, and debug while retaining session state across reconnects. Each interactive session uses a runtime session ID and shell ID for resume; up to 10 concurrent shells are supported per runtime.

🔐 Researchers disclosed a major data exposure at Moltbook on January 31, 2026, revealing 35,000 emails and 1.5 million agent API tokens across 770,000 agents. Private messages contained plaintext third-party credentials, including OpenAI API keys, creating what the article calls a toxic combination — cross-app permissions that compound risk. The piece urges shifting review from single apps to the bridges between them and highlights procedural controls and dynamic SaaS security platforms like Reco to monitor runtime trust relationships and revoke risky tokens before exfiltration.

🔒 Kaspersky has extended Kaspersky Container Security with support for the OpenAI API, allowing organizations to connect local or third‑party large language models that implement that API. The integrated AI assistant analyzes uploaded container images, describes their contents and behavior, performs independent risk assessments, and suggests mitigations to speed investigations and decision-making. The update also brings single sign‑on and multi‑domain Active Directory support, faster image scanning, and enhanced security policy capabilities to the Kaspersky Cloud Workload Security suite.

🔒 Amazon Bedrock now supports AWS PrivateLink for the bedrock-mantle endpoint, enabling private network access to OpenAI API-compatible service endpoints. The bedrock-mantle endpoint is powered by Project Mantle, a distributed inference engine that simplifies model onboarding and delivers serverless, high-performance inference with QoS controls and higher default quotas. This expansion gives enterprises a private connectivity option across multiple AWS Regions for building and scaling generative AI applications.

🚀 Amazon Bedrock now exposes an OpenAI-compatible Responses API on new service endpoints, enabling asynchronous inference for long-running workloads, streaming and non-streaming modes, and automatic stateful conversation reconstruction so developers no longer must resend full histories. The endpoints provide Chat Completions with reasoning-effort support for models served by Mantle, Amazon’s distributed inference engine. Integration requires only a base URL change for OpenAI SDK–compatible code, and support starts today for OpenAI’s GPT OSS 20B and 120B models, with additional models coming soon.

🔒 Microsoft’s Detection and Response Team (DART) detailed a novel .NET backdoor called SesameOp that leverages the OpenAI Assistants API as a covert command-and-control channel. Discovered in July 2025 during a prolonged intrusion, the implant uses a loader (Netapi64.dll) and an OpenAIAgent.Netapi64 component to fetch encrypted commands and return execution results via the API. The DLL is heavily obfuscated with Eazfuscator.NET and is injected at runtime using .NET AppDomainManager injection for stealth and persistence.

🔐 Microsoft security researchers identified a new backdoor, SesameOp, which abuses the OpenAI Assistants API as a covert command-and-control channel. Discovered during a July 2025 investigation, the backdoor retrieves compressed, encrypted commands via the API, decrypts and executes them, and returns encrypted exfiltration through the same channel. Microsoft and OpenAI disabled the abused account and key; recommended mitigations include auditing firewall logs, enabling tamper protection, and configuring endpoint detection in block mode.