Amodei Tells CBS the Industry Lied About AI Risk and Pacing Beats a Pause

Classy aggregates publicly available material; we did not conduct a private interview.

Watch the public conversation

What changed

On 13 September 2026, Anthropic CEO Dario Amodei sat for a 23 minute extended interview with CBS Sunday Morning correspondent Jo Ling Kent at Anthropic headquarters in San Francisco. The segment aired as labs face mounting pressure to prove safety work keeps pace with frontier model releases.

Amodei said exponential AI progress is a warning sign, not a reason to panic or halt research entirely. He argued the industry lied for too long about risks, and that pacing the frontier now requires verifiable oversight rather than voluntary blog posts.

Why it matters

The interview landed the same weekend Amodei published his We Must Pace the Frontier essay and Anthropic committed to employee level access for embedded third party evaluators. Sam Altman and Demis Hassabis publicly backed pacing and evaluator access within days, signaling the debate shifted from whether to slow down to how verification gets funded and standardized.

For enterprise buyers, the CBS exchange clarifies that safety procurement will increasingly ask vendors to document evaluator access, incident reporting, and training oversight rather than marketing safety scores alone.

Who is affected

Frontier lab policy teams must translate CEO pledges into contracts evaluators can audit. Regulators and standards bodies gain public cover for FINRA style oversight proposals Hassabis outlined in July. Risk officers at banks, health systems, and defense contractors should treat pacing rhetoric as a cue to delay irreversible autonomy deployments until evaluator reports exist.

What to do next

Brief your governance committee on Amodei's three part pacing plan: embedded evaluators, coordinated safety standards, and limits on unchecked capability jumps. Map which vendor relationships already include audit rights and which rely on trust me blog posts.

What to watch

Track whether OpenAI, Google, and Anthropic publish joint standards body charter language after September 15 industry reports of active talks. The next verifiable signal is a named oversight board with funding commitments, not another CEO thread on X.

Question (Jo Ling Kent): Could AI kill us in the near future?

Amodei (paraphrased from CBS): This is a warning sign. We need to slow down, develop carefully, and ensure safety measures keep pace with capability.

Question (Jo Ling Kent): Does the industry need to stop entirely?

Amodei (paraphrased from CBS): No panic and no complete halt. Pacing and verification are the responsible path.

Sources

1. Primary. CBS News, Extended Interview: Dario Amodei (13 September 2026). Full video record for quotes and pacing plan discussion.
2. Primary. CBS News, Anthropic CEO: For too long the industry lied about AI risks (13 September 2026). Broadcast context and web exclusive link.
3. Secondary. PANews, Anthropic CEO on Slowing Down AI Development (13 September 2026). Corroborates timing with Amodei essay and Altman support.