Latest UK & Global AI News

In this 23 July 2026 published video, hosts Krystal Ball and Saagar Enjetif talk with Garrison Lovely about how an OpenAI model escaped containment into the real world and hacked a real company. The AI was tasked with solving a cybersecurity [...]

View Full Article

President Donald Trump is having governors and electricity companies join a pledge aimed at shielding U.S. consumers from higher utility bills because data centers. This move highlights the controversy surrounding AI development before the November midterm elections. Trump initially announced...

View Full Article

Part 1: From one support bot to a framework At Grab, AI agents have evolved from interesting team prototypes into production services used every day by millions of merchants, drivers, and consumers. Today, more than 500 services run on our internal agent framework, over 50 Model Context Protocol (MCP) servers are registered on our remote MCP framework, and a single Large Language Model (LLM) gateway fronts every model call across the company, handling billions of tokens each month. None of this was designed up front. It began as the plumbing behind one internal support bot, which then expanded because the same problems kept resurfacing for every team trying to ship an agent. This series tells the story of what the platform eventually became. This Part 1 of the blog focuses on the beginning: the architecture of our AI support bot, the specific pain points we hit while scaling and iterating on it, and how each of those failures became a core building block in the framework we now call LLM-Kit . The bot that started it Imagine you have a question for the Technical Infrastructure (Tech Infra) team - the engineers who run the cloud platforms, databases, developer tooling, and AI infrastructure behind Grab’s ecosystem. Instead of immediately paging an on-call engineer, a bot first triages the request, checks the team’s documentation, runbooks, and past Slack threads, and tries to answer directly in the thread. If it still cannot resolve the issue, it routes the ticket to the right human, with the relevant context already attached. That is what we built with the Tech Infra Support Bot. In the first half of 2023, Tech Infra handled thousands of support tickets, many of them repeated questions that had already been answered somewhere internally. Before LLMs, the bot’s role was mainly operational; performing tasks like helping track acknowledgments and response times for on-call engineers. With the arrival of GPT-4-32k, we evolved it into a GPT-powered Level-0 support layer that could answer documented questions before a human needed to be paged. The first production version was a Go service organized around two planes: A reasoning plane . At Level-0, it was a single-agent loop. It takes the user’s question, decides which tools to call, executes those calls, feeds the results back into the prompt, and returns an answer. The default model at the time was gpt-4.1; today, we have evolved to the latest reasoning models. A tool plane . The tools provided the bot’s core working context. Retrieval flowed through Glean, which covered Confluence, TechDocs , internal drives, and Jira. Other tools handled log search through Kibana, GitLab runbook and file access, Slack conversation search, and a small set of Hypertext Transfer Protocol (HTTP) plugins. In the first version, tools and prompts were defined in per-channel JavaScript Object Notation (JSON) configs and resolved at request time. As models became more capable, we later standardized the tool set across channels. A trimmed version of that tool config looked like this: "agent_plugins" : [ { "name" : "glean_search" , "type" : "common" , "metadata" : { "wiki_space_collection" : [ "..." ]}}, { "name" : "runbook_search" , "type" : "common" }, { "name" : "gitlab_runbook_reader" , "type" : "common" }, { "name" : "gitlab_read_file" , "type" : "common" }, { "name" : "kibana_log_search" , "type" : "common" , "metadata" : { "index" : "k8s*" }}, { "name" : "slack_conversation_tool" , "type" : "common" } ] It worked, but it taught us, the hard way, why a demo agent is not a production agent. What it takes to scale and improve quickly As we worked on improving the agent, we kept running into the same kinds of friction. Over time, those pain points formed clear patterns, and they were the same ones we saw other teams run into as well. Vibe check is not an evaluation strategy . The bot had a base prompt, and each Slack channel could configure its own prompt, tools, and documentation filters. But the workflow was essentially: configure it, ship it, and hope it reduced toil. There were no real evaluations, just optimism that it would work. Fast model and provider switching is essential . The AI landscape moves incredibly fast: a new state-of-the-art (SOTA) model appears on Tuesday, and a highly efficient open-source alternative shows up on Thursday. Switching providers should not feel like open-heart surgery. A unified Software Development Kit (SDK) and an LLM API gateway remove the need to refactor payload schemas, rewrite error handling, or integrate each provider from scratch. If moving from OpenAI to Anthropic, or routing to an open-source model endpoint, takes more than a few config changes, technical debt is already slowing you down. Observability cannot be an afterthought . When an answer was wrong, figuring out “why” meant grepping logs across three separate systems: the agent workflow, the tool calls, and the model call. There was no shared trace tying them together. That level of friction is survivable for an internal tool; it is unacceptable for a customer-facing agent. Everything around the agent took longer than the agent itself . Auth (OIDC), secrets management (Vault), per-environment config, vector database integration, LLM tracing, health probes, and metrics were not agent-specific problems. However, they all had to be solved before anything could be shipped. The reasoning loop took a whole afternoon. The production wrapper took two weeks. The pattern was clear: the hard part of building an agent was not the agent itself, but everything around it that had to be in place before it could safely run in front of users. So we began pulling those shared components out of the bot and consolidating them into a unified framework. Extracting the framework: LLM-Kit LLM-Kit emerged when we stopped solving these problems service by service and started solving them once, centrally. It is intentionally not a new agent abstraction or a Domain-Specific Language (DSL). Instead, it is a curated set of integrations and scaffolding built around Grab’s existing infrastructure, pipelines, secret management, and observability. Just as importantly, we chose to build a framework rather than a heavy centralized platform. In a space evolving this quickly, a platform would have locked teams into rigid assumptions that would soon become outdated. A framework let us meet developers where they already were: standardizing the plumbing while preserving the freedom to iterate quickly. Looking back, that was the right first choice. Each part of LLM-Kit is a direct response to one of the failures described above. We first wrote about LLM-Kit’s structure and code architecture in a 2024 blog post . Two years and a few hundred agents later, the overall shape is still recognizable, but almost every underlying layer has changed. Poetry was replaced by uv ; we standardized on the OpenTelemetry stack; LangChain evolved into LangGraph and Deep Agents; and some tools moved onto our MCP framework. It starts with a template . The entry point is a user interface (UI) form. An engineer fills in an application name and a few details, and gets back a GitLab repository with the production wrapper already assembled. Under the hood the template stamps out a full FastAPI service: / ├── app/ │ ├── server.py # FastAPI app factory: mounts routes + middleware, boots OTel + statsd │ ├── agents/ │ │ ├── simple_react_agent.py # a single-agent LangGraph ReAct loop (agent <-> tools) │ │ ├── mcp_react_agent.py # the same loop, but tools are pulled from remote MCP servers │ │ └── simple_react_agent.png # auto-exported graph diagram (generated in dev) │ ├── routes/ │ │ ├── api.py # router aggregator │ │ ├── health_check.py # liveness/readiness probe │ │ ├── oidc.py # OIDC login/callback (skipped in proxy-auth mode) │ │ └── evalshub_eval.py # runs ROUGE / BLEU / LLM-as-judge evals on the agent │ ├── core/config.py # AppConfig (pydantic-settings) + INI/secret parsing │ ├── tools/word_length_tool.py # an example tool to copy from │ ├── utils/prompts.py # prompt/message assembly helpers │ └── storage/connection.py # Postgres + pgvector engine and connection pooling ├── sdk/ # a generated, typed client SDK (protobuf) other services import ├── configs/ │ ├── dev.ini / stg.ini / prd.ini # one config per environment │ └── secret.ini.example # secret template; real values resolve from Vault at deploy ├── databases/postgresql/ # SQL migrations (pgvector extension bootstrapped for you) ├── scripts/ │ ├── db.py / db.sh # migration runner │ └── gunicorn_conf.py # production server/worker config ├── tests/ │ ├── unit_tests/ # starter unit tests (e.g. the health check) │ └── evalshub_evaluation/ # golden test cases the eval route runs against ├── Dockerfile # multi-stage, distroless ├── Makefile # setup / run / test / lint targets ├── pyproject.toml # uv build backend + pinned deps └── .pre-commit-config.yaml Three things are worth pulling out of that tree: app/agents/ is the part you actually own . You get two working agents to fork from rather than a blank file: simple_react_agent.py is a single-agent LangGraph ReAct loop, and mcp_react_agent.py is the same loop wired to pull its tools from remote MCP servers. Both compile to a LangGraph StateGraph with a retry policy and a 30-second per-step timeout, and in dev the graph is auto-exported as a diagram. This is a real step up from the bare LangChain agent initialization we scaffolded in 2024. app/routes/evalshub_eval.py ships evals on day one . The template comes with an endpoint that runs Recall-Oriented Understudy for Gisting Evaluation (ROUGE), Bilingual Evaluation Understudy (BLEU), and LLM-as-judge evaluators over a set of golden test cases in tests/evalshub_evaluation/ . The thing we most wished the support bot had, is now in the box before a builder writes a line of their own logic. Everything else is the production wrapper . core/config.py , storage/ , configs/ , databases/ , scripts/ , the distroless Dockerfile , and the pyproject.toml (now uv , not the Poetry we used in 2024) are the auth, secrets, persistence, packaging, and deploy plumbing that every service needs and that no team should have to write from scratch. The day-one wiring that used to take two weeks or more now takes about an hour. The rest of this section is what “pre-wired” means, layer by layer. Config and secrets are solved once . Apps declare environment configs as initialization (INI) files with secret interpolation, so secrets resolve from Vault at boot, and a single secret.ini.example is enough to run any LLM-Kit app locally: [CONFIG] GRABGPT_API_KEY=${SECRET:GRABGPT_API_KEY} OTEL_EXPORTER_OTLP_ENDPOINT= POSTGRES_POOL_RECYCLE=1800 Model access behind one resolver . Every model call goes through the GrabGPT Gateway, which is OpenAI-compatible. LLM-Kit’s job is just to resolve the right endpoint (per environment, and per data tier) and inject the key so application code never hard-codes a provider again: from openai import OpenAI from llm_kit.grabgpt import resolve_grabgpt_base_url , resolve_grabgpt_api_key client = OpenAI ( base_url = resolve_grabgpt_base_url ( " prd " , " public " ), # provider chosen centrally api_key = resolve_grabgpt_api_key (), ) That one indirection is what later lets a platform team change which provider serves a model, configure fallback routing, set budgets, and manage cost attribution, without a single application touching its code. Tracing wired in, not bolted on . A single instrumentor auto-instruments FastAPI, outbound HTTP, LangChain, and MCP, and stamps every span with Kubernetes resource attributes (pod, namespace, image, service version). Structured logs auto-inject the trace and span IDs, so logs and traces correlate in Grafana/Kibana for free: exporter = OTLPSpanExporter ( endpoint = app_config . otel_exporter_otlp_endpoint ) OTELInstrumentor ( exporter = exporter , excluded_urls = [ " health_check " ]). instrument_app ( app ) The three systems, no shared trace problem turns into one end-to-end trace across every LLM call, tool call, and retrieval step. Tools can be exposed through MCP servers built on our MCP framework . Instead of hardwiring a large set of tool functions inside the agent process, the agent connects to MCP servers and discovers their tools at runtime. That means adding a new capability can be as simple as registering an MCP server, rather than redeploying the agent. client = MultiServerMCPClient ({ " mcp-gitlab-remote " : { " transport " : " streamable_http " , " url " : " /mcp/ " , " headers " : { " Authorization " : " Bearer " }, } }) tools = await client . get_tools () # schema negotiated, no redeploy An agent is just another service in the ecosystem, with gRPC on both sides. Most of Grab’s backend communicates over gRPC, and agents are rarely standalone; other services call them, and they in turn call other internal services. The template is designed to support both directions. On the serving side, the scaffold includes a Protocol Buffers (protobuf) contract ( sdk/.../.proto , with a sample Hello remote procedure call (RPC)) and a generated, typed client SDK package that other teams import to call your agent without hand-writing HTTP. make gen-proto regenerates the Python stubs from the .proto , and a gen-proto-check Continuous Integration (CI) step fails the build if the committed stubs drift from the contract. A gRPC server runs alongside FastAPI (default port 8087 , multi-worker-safe via SO_REUSEPORT ) and ships a standard gRPC health service out of the box: $ grpcurl -plaintext localhost:8087 grpc.health.v1.Health/Check On the calling side, LLM-Kit ships a channel provider so an agent never hardcodes an address. The auto provider tries Istio, then Consul, then a static fallback, health-checks the channel it selects, and runs a background monitor that re-selects after a few consecutive failures: from llm_kit.grpc.channel_providers.auto import ( AutoGrpcChannelProvider , AutoGrpcChannelProviderConfig , ) provider = AutoGrpcChannelProvider ( logger , AutoGrpcChannelProviderConfig ( client_name = " my-agent " , service_key = " some-internal-service " , # resolved via Istio / Consul enable_istio = True , enable_consul = True , )) channel = provider . get_channel () # first healthy channel, auto-reselected on failure stub = SomeServiceStub ( channel ) This is the less glamorous side of being production-ready . Before an agent can deliver value, it needs to both accept calls from and make calls to the rest of the company’s services using the same transport the broader system already relies on. What’s next LLM-Kit solved building and shipping one agent. At 500 agents, the problems were no longer framework problems. They were platform problems: who can change which model everyone calls, how one team safely reuses another team’s tools, and how you know an agent got better and not just different after a prompt change. We built three answers for that layer: the GrabGPT Gateway, a remote MCP framework, and an evals platform. Part 2 starts with the gateway — one endpoint, five providers, and what it takes to make “swap the model” a configuration change instead of an incident. Join us Grab is Southeast Asia’s leading superapp, serving over 900 cities across eight countries (Cambodia, Indonesia, Malaysia, Myanmar, the Philippines, Singapore, Thailand, and Vietnam). Through a single platform, millions of users access mobility, delivery, and digital financial services, including ride-hailing, food delivery, payments, lending, and digital banking via GXS Bank and GXBank. Founded in 2012, Grab’s mission is to drive Southeast Asia forward by creating economic empowerment for everyone while delivering sustainable financial performance and positive social impact. Powered by technology and driven by heart, our mission is to drive Southeast Asia forward by creating economic empowerment for everyone. If this mission speaks to you, join our team today!

View Full Article

The Child Mind Institute will combine clinical assessments, digital journals and de-identified chat histories to develop markers for evaluating young people’s mental states. The Child Mind Institute has begun a one-year research initiative, supported by a grant from the OpenAI Foundation , to examine youth mental health during interactions with AI chatbots in the United States. Rather than testing a particular chatbot or presenting evidence about its effects, the project will build the research infrastructure needed to measure young people’s mental states while they use conversational AI and over time. The aim is to identify valid markers that could inform the design and testing of safer AI systems for children. Researchers plan to combine clinical assessments, digital journals, de-identified AI chat histories, real-time behavioral measures and existing youth mental health datasets. The initial phase will identify signals that could be used to assess and monitor mental health alongside young people’s use of AI tools. The institute says young people are turning to general-purpose chatbots, digital companions and products described as “therapy bots” for support amid a shortage of qualified mental health professionals, stigma and misinformation. The initiative is designed to close an evidence gap around tools that young people are already using, without treating AI chatbots as a substitute for professional support. Harold S. Koplewicz, President and Medical Director at the Child Mind Institute, places trained clinicians, research and safeguards at the center of any future role for the technology. “It is our belief that with appropriate safeguards and evidence, digital tools may complement care from trained clinicians,” Koplewicz says. “We are excited about filling a gap that currently exists in the research of AI tools and to work toward creating a safer online experience for youth around the globe.” Independent research with privacy protections The Child Mind Institute will develop and conduct the research independently. The OpenAI Foundation is providing financial support, but the initiative will be solely executed by the institute. The project will be co-designed by Principal Investigators Gregory Kiar, Arno Klein and Michael P. Milham, Chief Science Officer at the Child Mind Institute. Their combined areas of expertise include computational methods, digital measurement, clinical science and youth mental health. Milham says the research will consider both the risks associated with AI use and the conditions under which conversational tools could support existing mental health services. “It is critical to better understand the relationship between AI use and youth mental health, and to examine whether, and under what conditions, AI tools can strengthen evidence-based care, support clinician training, and expand access to high-quality mental health services,” he explains. The institute says the study will use informed consent, ethical oversight, privacy protections and data governance measures to protect participants. AI chat histories used by researchers will be de-identified. The Child Mind Institute also plans to share data from the project to support further research.

View Full Article

Drug development is among the slowest, most failure-prone processes in modern science, with about 90% of drug candidates never reaching the market. Today, artificial intelligence methods have accelerated the first step—plucking promising molecules out of endless possibilities—but countless challenges remain. A successful drug must be not only safe and effective but also able to bypass the body's defenses and reach the right target.

View Full Article

View Full Article

View Full Article

The best local & breaking news source in the US, featuring local weather, alerts, deals, events and more.

View Full Article

New authorisation framework under the Telecommunications Act, 2023 mandates domestic storage of telecom network data, expands government oversight and places rollout obligations on infrastructure providers

View Full Article

View Full Article

View Full Article

LivePerson Inc. announced that its merger with SoundHound AI, Inc. has received all necessary foreign investment approvals. The final clearance came from the Bulgarian authority on July 20, 2026, following earlier approvals from Italy, Canada, Germany, and the United Kingdom. While all regulatory conditions are now met, the merger still requires LivePerson stockholder approval to be consummated.

View Full Article

View Full Article

The enterprise generative ai market size is projected to grow from $5245.12 million in 2026 to $59252.2 million by 2034, at a CAGR of 35.4% during the forecast period 2026-2034.

View Full Article

View Full Article

Vodafone autonomous antenna trial in Albania pairs a HUMAX Networks robotic arm with self-organizing AI to physically rotate and tilt 4G and 5G antennas in under 30 minutes, replacing engineer site visits that previously took weeks. The system adjusts coverage between a shopping center and residential areas without human intervention.

View Full Article

U.S. private equity firms like Blackstone and Apollo see shares fall over 20% as the AI boom benefits investment banks, while missed AI investments and rising redemptions weigh on the sector.

View Full Article

Christopher Nolan’s The Odyssey has made $454m worldwide since its debut.

View Full Article

AI infrastructure buildouts are moving so fast that plans made just months ago are already obsolete, forcing hardware makers to rewrite how they design, source and ship the systems powering the next generation of AI factories. The shift from single-shot retrieval-augmented generation toward agentic AI has reshaped what enterprises need from their compute stacks almost [...] The post AI infrastructure demand is outrunning even the boldest supply chain playbooks appeared first on SiliconANGLE .

View Full Article

OpenAI’s models autonomously hacked a tech startup. It signals a seismic shift in cybersecurity The Conversation OpenAI says its AI went rogue and launched 'unprecedented' cyber-attack BBC For years tech experts imagined AI breaking free. Now they have to stop it. The Washington Post AI agent went rogue and hacked startup by itself, OpenAI reveals The Guardian OpenAI and Hugging Face partner to address security incident during model evaluation OpenAI

View Full Article

MANILA - South Korea and the Association of Southeast Asian Nations (ASEAN) on Wednesday reaffirmed their commitment to advancing the implementation of their Comprehensive Strategic Partnership vision and introduced new cooperation initiatives in artificial intelligence (AI) and cultural and creative industries,.

View Full Article

(Bloomberg) -- China should regulate quant funds and the usage of artificial intelligence, Shanghai Securities News reported, citing investors and experts af...

View Full Article

Tesla’s record Q2 deliveries failed to translate into higher profits as price cuts, falling regulatory-credit revenue and Elon Musk’s aggressive AI spending squeezed margins and pushed the EV maker into negative free cash flow

View Full Article

Woking, UK - Launching on 23th July, Votito is a product management platform built for a shift the software industry: as AI tools make software faster to write, the constraint is no longer building things, it is knowing which things deserve to be built.

View Full Article

View Full Article